From 7d640365b63bbd7ee244d1490efa0f1a3f12d544 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Sun, 27 Sep 2026 19:57:04 +0800 Subject: [PATCH 001/162] feat(subagent): share fixed model-selection tools across sessions --- ...authorized-subagent-model-routes.i18n.yaml | 12 +- ...4-user-authorized-subagent-model-routes.md | 12 +- ...ser-authorized-subagent-model-routes.zh.md | 12 +- apps/cli/composition.md | 3 + .../tests/profiles/sdk/keyless-smoke.e2e.ts | 2 +- apps/cli/tests/web-agent-presets.e2e.ts | 12 +- docs/config-catalog.md | 2 +- docs/config-catalog.zh.md | 2 +- docs/event-producer-consumer.md | 2 +- docs/event-producer-consumer.zh.md | 2 +- docs/persistence-catalog.md | 6 +- docs/persistence-catalog.zh.md | 6 +- docs/tool-catalog.i18n.yaml | 12 +- docs/tool-catalog.md | 22 ++- docs/tool-catalog.zh.md | 22 ++- packages/bundle/base/cordis.patch.yml | 8 +- packages/bundle/web-app/cordis.patch.yml | 3 + .../bundle/web-app/presets/cordis.patch.yml | 2 + packages/bundle/web-app/presets/ptc.patch.yml | 2 + .../bundle/web-app/presets/standard.patch.yml | 2 + .../subagent/tool-subagent/README.i18n.yaml | 28 ++-- packages/subagent/tool-subagent/README.md | 17 +-- packages/subagent/tool-subagent/README.zh.md | 17 +-- packages/subagent/tool-subagent/package.json | 7 +- packages/subagent/tool-subagent/src/index.ts | 81 ++++------- .../subagent/tool-subagent/src/list-models.ts | 47 ++++-- .../src/model-selection-state.ts | 18 +-- .../tool-subagent/src/model-selection.ts | 11 +- .../subagent/tool-subagent/tests/harness.ts | 5 +- .../tool-subagent/tests/list-models.spec.ts | 52 +++++-- .../tests/model-selection-settings.spec.ts | 112 +++++++++------ .../tests/model-selection.spec.ts | 59 +++++++- .../tool-subagent/tests/tool-subagent.spec.ts | 9 +- .../subagent/tool-subagent/tsdown.config.ts | 1 + scripts/check-workspace-constraints.ts | 5 +- scripts/gen-tool-catalog.ts | 6 +- .../advanced/result.json | 6 + .../advanced/session.1.v4.jsonl | 2 +- .../advanced/session.2.v4.jsonl | 2 +- .../advanced/session.v4.jsonl | 4 +- .../scheduler-recovery/session.v4.jsonl | 2 +- .../sdk/bash-tool/tool-schemas.expected.json | 45 +++++- .../notifications.expected.jsonl | 6 +- .../sdk/dynamic-tool-updates/session.v4.jsonl | 6 +- snapshots/sdk/persistent-tools/cordis.yml | 4 + .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.expected.json | 73 +++++++++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.expected.json | 18 ++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.1.expected.json | 45 +++++- .../sdk/text-turn/tool-schemas.expected.json | 45 +++++- .../notifications.expected.jsonl | 4 +- .../sdk/tool-error-details/session.v4.jsonl | 4 +- .../system-prompt.expected.md | 24 +++- .../tool-schemas.expected.json | 45 +++++- .../advanced-toolchain/session.v4.jsonl | 8 +- .../system-prompt.expected.md | 24 +++- .../tool-schemas.expected.json | 90 +++++++++++- .../tool-schemas.expected.json | 90 +++++++++++- .../tool-schemas.expected.json | 45 +++++- .../both-mode-turn/system-prompt.expected.md | 24 +++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../cordis.snapshot.yml | 4 +- .../compaction-output-reserve/cordis.yml | 2 +- .../session.v4.jsonl | 2 +- .../tool-schemas.expected.json | 90 +++++++++++- .../cordis.snapshot.yml | 4 +- .../compaction-summary-headroom/cordis.yml | 2 +- .../session.v4.jsonl | 2 +- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../session.v4.jsonl | 6 +- .../dynamic-tool-updates/session.v4.jsonl | 6 +- .../tool-schemas.expected.json | 135 +++++++++++++++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../lsp-definition/tool-schemas.expected.json | 45 +++++- .../system-prompt.expected.md | 24 +++- .../mcp-resources/tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../office-skills/tool-schemas.expected.json | 45 +++++- .../plugin-manager-mcp/session.v4.jsonl | 4 +- .../tool-schemas.expected.json | 90 +++++++++++- .../plugin-manager/tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../provider-cwd/tool-schemas.expected.json | 45 +++++- .../system-prompt.expected.md | 24 +++- .../system-prompt.expected.md | 24 +++- .../ptc-python-turn/system-prompt.expected.md | 25 +++- .../ptc-read-image/system-prompt.expected.md | 24 +++- .../ptc-turn/system-prompt.expected.md | 24 +++- .../tool-schemas.expected.json | 45 +++++- .../ralph-loop/tool-schemas.1.expected.json | 45 +++++- .../ralph-loop/tool-schemas.2.expected.json | 45 +++++- .../ralph-loop/tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.1.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../team-targets/tool-schemas.expected.json | 17 +++ .../text-turn/tool-schemas.expected.json | 45 +++++- .../web-fetch/tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../tool-schemas.expected.json | 45 +++++- .../worktree/tool-schemas.expected.json | 45 +++++- tsconfig.base.json | 1 + 121 files changed, 3226 insertions(+), 411 deletions(-) diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index d4f615792e..c5c1287e45 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -9,14 +9,14 @@ en: 3238dfb7db7dfdd8 zh: ec3bdd9a9da0a321 /agent-note-user-authorized-subagent-model-routes/decision: - en: 0166e452a774ade5 - zh: 747061b9424b3610 + en: 499721e05ad15d9c + zh: 2bcf5ade8910da8e /agent-note-user-authorized-subagent-model-routes/alternatives-considered: en: e41c17226190ea8b zh: d3b1395fafe75f01 /agent-note-user-authorized-subagent-model-routes/consequences: - en: f1927d31ffc1dee9 - zh: b0fed04b80043c71 + en: 29ca812656549a01 + zh: aacd814d7d649dbf /agent-note-user-authorized-subagent-model-routes/related-decisions: - en: 595a9485d91592c6 - zh: 7b32611ea380498c + en: 17fdfabedb673935 + zh: 496f4f85622d7d15 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index e7f3b4fcae..3e2ab42e5a 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -12,11 +12,11 @@ Registering an LLM adapter makes its routes reachable, but does not authorize an The Host-owned `subagent-model-selection` settings section stores an explicit `enabled` switch and `allowedModels`, an array of exact `{ provider, model }` routes. Enabling requires at least one route; disabling may retain the selected routes for later reuse. The Plugins settings card reads the live adapter directory through `session/modelCatalog`, lets the user stage the switch and routes, and saves both fields in one revision-fenced settings mutation. It stores no adapter-owned display names, descriptions, or reasoning-effort metadata. A stored or staged route absent from the current directory remains visible as unavailable and removable; a provider-local catalog failure does not block other providers or erase saved authorization or an unsaved selection. A connection reset discards the draft because namespace revisions are comparable only within one Host process. -A newly composed top-level Session snapshots the route list in `subagent/model-selection-policy` when the setting is enabled, before its model-selectable definitions can reach a request. Event presence means selection was enabled; the event does not store the global switch. Child Sessions inherit that exact list from their live parent, and resumed Sessions use the recorded event instead of current settings. Settings changes therefore affect only subsequently composed top-level Sessions, while a legacy Session without the event remains disabled, including an explicitly empty restored Session. +A newly composed top-level Session snapshots the route list in `subagent/model-selection-policy` when the setting is enabled, before its first model request. Event presence means selection was enabled; the event does not store the global switch. Child Sessions inherit that exact list from their live parent, and resumed Sessions use the recorded event instead of current settings. Settings changes therefore affect only subsequently composed top-level Sessions. Without the event, including an explicitly empty restored Session, explicit selection remains restricted to the current parent route. -The fixed `list_subagent_models` schema does not enumerate the policy. At call time, provider and model listings are the intersection of the Session route list and the adapter's live advertised directory. An exact provider/model lookup first requires authorization, then resolves the adapter-owned model metadata and all advertised reasoning efforts. The delegation executor independently rejects any explicit provider, model, or effort selection whose effective provider/model route is outside the Session list before `resolveCallConfig()` validates adapter availability and effort support. A call that supplies no selection field retains configured or inherited routing because the model made no route choice. +The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and resolves their allowed exact model ids, including ids absent from advisory catalogs. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts. Without a recorded policy, discovery uses the parent Agent’s current effective route. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. -Model selection has no unrestricted static mode. The default-off Host setting is the only authority, and an enabled Session always carries an exact allowlist. The primary spawn tool reads that setting; the shipped fork tool still exposes no route selection so inherited conversation prefixes remain eligible for provider-side KV Cache reuse. +Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. The primary spawn tool captures the setting once, while all capable delegation tools and shared discovery read the same Session policy. Fork defaults preserve the parent route, and its description warns about prefix recomputation when that route changes. ## Alternatives considered @@ -33,11 +33,11 @@ Model selection has no unrestricted static mode. The default-off Host setting is ## Consequences - New adapter registrations and newly advertised models do not expand user authorization. -- Adapter removals or catalog failures can reduce what discovery currently lists without deleting the saved route decision; an exact authorized route remains usable when its adapter accepts it even if the advisory catalog omits it. +- Adapter removal or exact-model resolution failure can make discovery unavailable without deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. - The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. - The policy event is log-only and is appended while an Agent is composed, before either SDK begins its run subscription. Shipped SDK profiles do not enable this Web-owned preference, so the event changes neither SDK's expected notifications or persisted-session output; package restore tests own its durable projection instead of fabricating an SDK composition solely to emit it. -- Unit coverage pins settings validation, malformed durable values, Session sampling and inheritance, discovery intersection, executor denial, live UI catalog invalidation, staged-route retention, connection-generation invalidation, staged whole-array writes, stale-revision rejection, and retry after scoped installation failure. The assembled Web scenario pins the real settings document and Plugins card flow. +- Unit coverage pins settings validation, malformed durable values, Session sampling and inheritance, shared definitions, parent-route defaults, discovery restrictions, executor denial, and the settings UI’s staged writes and revision checks. The assembled Web scenario pins the real settings document and Plugins card flow. ## Related decisions -The route arguments, adapter preflight, discovery tool, and fork cache restriction remain owned by [model-selected subagent routes](../../../../packages/subagent/tool-subagent/README.md). +The route arguments, adapter preflight, discovery tool, and fork cache costs remain owned by [model-selected subagent routes](../../../../packages/subagent/tool-subagent/README.md). diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index f55f60de1b..d90bca9358 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -12,11 +12,11 @@ Status: implemented Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` 开关与 `allowedModels`，后者是由精确 `{ provider, model }` 路由组成的数组。启用时必须至少有一条路由；关闭时可以保留已选路由，供以后重新启用。Plugins 设置卡通过 `session/modelCatalog` 读取实时适配器目录，让用户暂存开关与路由，再在一次带 revision 限制的设置 mutation 中保存两个字段。它不保存适配器自有的显示名称、描述或推理强度元数据。当前目录中缺失的已存或暂存路由仍显示为不可用并允许移除；某个提供方的目录失败不会阻塞其他提供方，也不会清除已存授权或未保存选择。连接重置会丢弃草稿，因为 namespace revision 只能在同一个 Host 进程内比较。 -设置启用时，新组合的顶层 Session 会在模型可选定义进入请求之前，把路由列表快照记录为 `subagent/model-selection-policy`。事件存在就表示模型选择已启用；事件不保存全局开关。子 Session 从在线父级继承同一份精确列表，恢复的 Session 使用已记录事件而不是当前设置。因此，设置修改只影响之后组合的顶层 Session，而没有该事件的旧 Session 仍保持禁用，包括显式为空的恢复 Session。 +设置启用时，新组合的顶层 Session 在首个模型请求前将路由列表记录为 `subagent/model-selection-policy`。事件存在表示选择已启用；事件不保存全局开关。子 Session 从在线父级继承同一精确列表，恢复的 Session 使用已记录事件而非当前设置。因此设置变更只影响之后组合的顶层 Session。没有该事件时，包括显式为空的恢复 Session，显式选择仍仅限当前父级路由。 -固定的 `list_subagent_models` schema 不会枚举该策略。调用时，提供方和模型列表是 Session 路由列表与适配器实时公布目录的交集。精确 provider/model 查询先要求授权，再解析适配器自有的模型元数据和全部已公布推理强度。委派执行器还会独立拒绝任何生效 provider/model 路由不在 Session 列表内的显式提供方、模型或强度选择，然后才由 `resolveCallConfig()` 校验适配器可用性与强度支持。完全没有选择字段的调用保留配置或继承路由，因为模型没有作出路由选择。 +固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并解析允许的精确模型 ID，包括建议性目录未列出的 ID。精确查询先要求授权，再解析模型元数据及全部支持的推理强度。没有已记录策略时，发现使用父 Agent 当前生效的路由。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 -模型选择不再有无限制的静态模式。默认关闭的 Host 设置是唯一授权来源，启用的 Session 始终携带精确允许列表。主 spawn 工具读取该设置；随附 fork 工具仍不公开路由选择，使继承的对话前缀继续符合提供方侧 KV Cache 复用条件。 +模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。主 spawn 工具一次性记录设置，所有支持选择的委派工具和共享发现均读取同一 Session 策略。fork 默认保留父级路由，其描述会提示改变路由时的前缀重算成本。 ## Alternatives considered @@ -33,11 +33,11 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` ## Consequences - 新适配器注册和新公布模型不会扩大用户授权。 -- 适配器移除或目录失败可以减少发现当前列出的内容，但不会删除已存路由决定；即使建议性目录省略某条精确已授权路由，只要适配器接受它，该路由仍然可用。 +- 适配器移除或精确模型解析失败可能使发现不可用，但不会删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 - 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。 - 策略事件仅存在于日志，并在 Agent 组合期间、两套 SDK 开始订阅运行前追加。随附 SDK profile 不启用这项 Web 自有偏好，因此该事件不会改变任一 SDK 的预期通知或持久 Session 输出；其持久投影由包级恢复测试负责，不会为了发出该事件而虚构 SDK 组合。 -- 单元覆盖固定设置校验、异常持久值、Session 取样与继承、发现交集、执行器拒绝、UI 实时目录失效、暂存路由保留、连接换代失效、暂存后的整数组写入、陈旧 revision 拒绝，以及作用域安装失败后的重试。组装 Web 场景固定真实设置文档与 Plugins 设置卡流程。 +- 单元覆盖固定设置校验、异常持久值、Session 取样与继承、共享定义、父级路由默认值、发现限制、执行器拒绝，以及设置 UI 的暂存写入与 revision 检查。组装 Web 场景固定真实设置文档与 Plugins 设置卡流程。 ## Related decisions -路由参数、适配器预检、发现工具与 fork 缓存限制仍由[模型选择的 subagent 路由](../../../../packages/subagent/tool-subagent/README.zh.md)负责。 +路由参数、适配器预检、发现工具与 fork 缓存成本仍由[模型选择的 subagent 路由](../../../../packages/subagent/tool-subagent/README.zh.md)负责。 diff --git a/apps/cli/composition.md b/apps/cli/composition.md index 9888602296..f2bae814f4 100644 --- a/apps/cli/composition.md +++ b/apps/cli/composition.md @@ -144,6 +144,8 @@ flowchart LR cfg --> plugin_dsh_base_subagent_fork_in_process plugin_dsh_base_tool_subagent_control["tool-subagent-control
@deepseek-ai/dsh-tool-subagent-control"] cfg --> plugin_dsh_base_tool_subagent_control + plugin_dsh_base_tool_subagent_list_models["tool-subagent-list-models
@deepseek-ai/dsh-tool-subagent/list-models"] + cfg --> plugin_dsh_base_tool_subagent_list_models plugin_dsh_base_tool_subagent_list_agents["tool-subagent-list-agents
@deepseek-ai/dsh-tool-subagent-control/list-agents"] cfg --> plugin_dsh_base_tool_subagent_list_agents plugin_dsh_base_tool_subagent["tool-subagent
@deepseek-ai/dsh-tool-subagent"] @@ -268,6 +270,7 @@ flowchart LR | `subagent-spawn-in-process` | `@deepseek-ai/dsh-subagent-spawn-in-process` | | `subagent-fork-in-process` | `@deepseek-ai/dsh-subagent-fork-in-process` | | `tool-subagent-control` | `@deepseek-ai/dsh-tool-subagent-control` | +| `tool-subagent-list-models` | `@deepseek-ai/dsh-tool-subagent/list-models` | | `tool-subagent-list-agents` | `@deepseek-ai/dsh-tool-subagent-control/list-agents` | | `tool-subagent` | `@deepseek-ai/dsh-tool-subagent` | | `tool-subagent-fork` | `@deepseek-ai/dsh-tool-subagent` | diff --git a/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts b/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts index 64c1c20ab3..f804cde4fa 100644 --- a/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts +++ b/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts @@ -178,7 +178,7 @@ describe('Python SDK dsh profile keyless smoke', () => { expect(modelRequests[0]?.max_tokens).toBe(1234) expect(toolNames).toEqual(expect.arrayContaining(['read', 'write', 'edit', 'web_fetch', 'web_search'])) expect(toolNames.includes('str_replace_editor')).toBe(editorEnabled) - expect(toolNames).not.toContain('list_subagent_models') + expect(toolNames).toContain('list_subagent_models') child.stdin.write(`${JSON.stringify({ jsonrpc: '2.0', id: 3, method: 'shutdown' })}\n`) const shutdown = await waitForLine(lines, value => value.id === 3, () => stderr) diff --git a/apps/cli/tests/web-agent-presets.e2e.ts b/apps/cli/tests/web-agent-presets.e2e.ts index db0d8e26e3..bb82f5349c 100644 --- a/apps/cli/tests/web-agent-presets.e2e.ts +++ b/apps/cli/tests/web-agent-presets.e2e.ts @@ -254,7 +254,7 @@ describe('the shipped Web composition', () => { // depend on ripgrep being present on the machine. expect(toolNames(ctx, handle.agent).filter(name => name !== 'glob' && name !== 'grep')).toEqual([ 'ask_user_question', 'bash', 'create_goal', 'edit', 'exit_plan_mode', - 'get_goal', 'interrupt_agent', 'job_kill', 'job_list', 'job_output', 'list_agents', 'present', 'read', 'read_image', + 'get_goal', 'interrupt_agent', 'job_kill', 'job_list', 'job_output', 'list_agents', 'list_subagent_models', 'present', 'read', 'read_image', 'schedule_create', 'schedule_delete', 'schedule_list', 'schedule_update', 'send_message', 'skill', 'subagent', 'subagent_fork', 'todo_write', 'update_goal', 'web_fetch', 'web_search', @@ -284,15 +284,19 @@ describe('the shipped Web composition', () => { setup: agentCtx => ctx.agentPresets.mount(agentCtx, 'standard').then(() => undefined), }) try { - expect(toolNames(ctx, disabled.agent)).not.toContain('list_subagent_models') - expect(toolParameterNames(ctx, disabled.agent, 'subagent')).not.toEqual(expect.arrayContaining([ + expect(toolNames(ctx, disabled.agent)).toContain('list_subagent_models') + expect(toolParameterNames(ctx, disabled.agent, 'subagent')).toEqual(expect.arrayContaining([ 'model', 'provider', 'reasoning_effort', ])) + expect(ctx.sessionProjections.stateOf(disabled.agent.session, 'subagentModelSelectionPolicy')).toBeNull() + expect(ctx.sessionProjections.stateOf(enabled.agent.session, 'subagentModelSelectionPolicy')).toEqual([ + { provider: 'deepseek-official', model: 'deepseek-v4-flash' }, + ]) expect(toolNames(ctx, enabled.agent)).toContain('list_subagent_models') expect(toolParameterNames(ctx, enabled.agent, 'subagent')).toEqual(expect.arrayContaining([ 'model', 'provider', 'reasoning_effort', ])) - expect(toolNames(ctx, disabled.agent)).not.toContain('list_subagent_models') + expect(toolNames(ctx, disabled.agent)).toContain('list_subagent_models') } finally { await ctx.settings.update(SUBAGENT_MODEL_SELECTION_SETTINGS_NAMESPACE, { enabled: false }) await enabled.dispose() diff --git a/docs/config-catalog.md b/docs/config-catalog.md index b091fa4da4..775ece7747 100644 --- a/docs/config-catalog.md +++ b/docs/config-catalog.md @@ -3916,7 +3916,7 @@ export interface Config { - `inject`: `tools` · `subagents` · `systemPrompt` · `sessionProjections` - `refs`: [`AgentOptions`](subsystems/core.md) -- `source`: [`packages/subagent/tool-subagent/src/index.ts:48`](../packages/subagent/tool-subagent/src/index.ts) +- `source`: [`packages/subagent/tool-subagent/src/index.ts:47`](../packages/subagent/tool-subagent/src/index.ts) ```ts config-catalog /** Config: which registered provider this tool delegates to, plus child defaults. */ diff --git a/docs/config-catalog.zh.md b/docs/config-catalog.zh.md index 399101c45c..370b740e74 100644 --- a/docs/config-catalog.zh.md +++ b/docs/config-catalog.zh.md @@ -3918,7 +3918,7 @@ export interface Config { - `inject`: `tools` · `subagents` · `systemPrompt` · `sessionProjections` - `refs`: [`AgentOptions`](subsystems/core.zh.md) -- `source`: [`packages/subagent/tool-subagent/src/index.ts:48`](../packages/subagent/tool-subagent/src/index.ts) +- `source`: [`packages/subagent/tool-subagent/src/index.ts:47`](../packages/subagent/tool-subagent/src/index.ts) ```ts config-catalog /** Config: which registered provider this tool delegates to, plus child defaults. */ diff --git a/docs/event-producer-consumer.md b/docs/event-producer-consumer.md index d070d6fd23..c36cd9801d 100644 --- a/docs/event-producer-consumer.md +++ b/docs/event-producer-consumer.md @@ -12,7 +12,7 @@ This matrix shows which packages dispatch each harness-owned event and which pac | `agent-preset/selected` | `emit` | [`packages/preset/agent-preset-registry/src/types.ts:69`](../packages/preset/agent-preset-registry/src/types.ts) | [`agent-preset-registry`](../packages/preset/agent-preset-registry) (`emit`) | `remotes` | | `agent/assistant-stream` | `emit` | [`packages/core/agent/src/runtime-types.ts:363`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`headless`](../packages/bundle/headless), `session-controller` | | `agent/created` | `serial` | [`packages/core/agent/src/runtime-types.ts:261`](../packages/core/agent/src/runtime-types.ts) | [`agent`](../packages/core/agent) (`serial`) | `agent-team`, `browser-use-runtime`, `claude-code-mods`, [`file-reference-local`](../packages/context/file-reference-local), [`goal`](../packages/goal/goal), [`goal-round-driver`](../packages/goal/goal-round-driver), [`hooks-claude-code`](../packages/experimental/hooks-claude-code), [`hooks-codex`](../packages/experimental/hooks-codex), [`loader-smoke`](../packages/test-support/loader-smoke), `session-controller`, `tool-agent-team`, [`tool-subagent`](../packages/subagent/tool-subagent) | -| `agent/disposed` | `emit` | [`packages/core/agent/src/runtime-types.ts:270`](../packages/core/agent/src/runtime-types.ts) | [`agent`](../packages/core/agent) (`events.dispatch`) | [`agent-loop`](../packages/core/agent-loop), `claude-code-mods`, [`file-reference-local`](../packages/context/file-reference-local), [`goal-round-driver`](../packages/goal/goal-round-driver), `session-controller`, [`subagent`](../packages/subagent/subagent), `tool-agent-team`, [`tool-subagent`](../packages/subagent/tool-subagent) | +| `agent/disposed` | `emit` | [`packages/core/agent/src/runtime-types.ts:270`](../packages/core/agent/src/runtime-types.ts) | [`agent`](../packages/core/agent) (`events.dispatch`) | [`agent-loop`](../packages/core/agent-loop), `claude-code-mods`, [`file-reference-local`](../packages/context/file-reference-local), [`goal-round-driver`](../packages/goal/goal-round-driver), `session-controller`, [`subagent`](../packages/subagent/subagent), `tool-agent-team` | | `agent/error` | `emit` | [`packages/core/agent/src/runtime-types.ts:393`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`acp`](../packages/acp/acp), [`deepseek-account`](../packages/credentials/deepseek-account), [`goal-round-driver`](../packages/goal/goal-round-driver), [`headless`](../packages/bundle/headless), `server`, `session-controller`, [`session-telemetry`](../packages/session/session-telemetry), [`subagent`](../packages/subagent/subagent) | | `agent/inbox/claimed` | `emit` | [`packages/core/agent/src/runtime-types.ts:299`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`acp`](../packages/acp/acp), [`goal-round-driver`](../packages/goal/goal-round-driver), [`subagent`](../packages/subagent/subagent), [`tool-jobs`](../packages/jobs/tool-jobs), [`user-questions`](../packages/interaction/user-questions) | | `agent/inbox/discarded` | `emit` | [`packages/core/agent/src/runtime-types.ts:307`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`goal-round-driver`](../packages/goal/goal-round-driver), [`subagent`](../packages/subagent/subagent), [`user-questions`](../packages/interaction/user-questions) | diff --git a/docs/event-producer-consumer.zh.md b/docs/event-producer-consumer.zh.md index 4d62fcfea2..2cc429dc63 100644 --- a/docs/event-producer-consumer.zh.md +++ b/docs/event-producer-consumer.zh.md @@ -14,7 +14,7 @@ | `agent-preset/selected` | `emit` | [`packages/preset/agent-preset-registry/src/types.ts:69`](../packages/preset/agent-preset-registry/src/types.ts) | [`agent-preset-registry`](../packages/preset/agent-preset-registry) (`emit`) | `remotes` | | `agent/assistant-stream` | `emit` | [`packages/core/agent/src/runtime-types.ts:363`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`headless`](../packages/bundle/headless), `session-controller` | | `agent/created` | `serial` | [`packages/core/agent/src/runtime-types.ts:261`](../packages/core/agent/src/runtime-types.ts) | [`agent`](../packages/core/agent) (`serial`) | `agent-team`, `browser-use-runtime`, `claude-code-mods`, [`file-reference-local`](../packages/context/file-reference-local), [`goal`](../packages/goal/goal), [`goal-round-driver`](../packages/goal/goal-round-driver), [`hooks-claude-code`](../packages/experimental/hooks-claude-code), [`hooks-codex`](../packages/experimental/hooks-codex), [`loader-smoke`](../packages/test-support/loader-smoke), `session-controller`, `tool-agent-team`, [`tool-subagent`](../packages/subagent/tool-subagent) | -| `agent/disposed` | `emit` | [`packages/core/agent/src/runtime-types.ts:270`](../packages/core/agent/src/runtime-types.ts) | [`agent`](../packages/core/agent) (`events.dispatch`) | [`agent-loop`](../packages/core/agent-loop), `claude-code-mods`, [`file-reference-local`](../packages/context/file-reference-local), [`goal-round-driver`](../packages/goal/goal-round-driver), `session-controller`, [`subagent`](../packages/subagent/subagent), `tool-agent-team`, [`tool-subagent`](../packages/subagent/tool-subagent) | +| `agent/disposed` | `emit` | [`packages/core/agent/src/runtime-types.ts:270`](../packages/core/agent/src/runtime-types.ts) | [`agent`](../packages/core/agent) (`events.dispatch`) | [`agent-loop`](../packages/core/agent-loop), `claude-code-mods`, [`file-reference-local`](../packages/context/file-reference-local), [`goal-round-driver`](../packages/goal/goal-round-driver), `session-controller`, [`subagent`](../packages/subagent/subagent), `tool-agent-team` | | `agent/error` | `emit` | [`packages/core/agent/src/runtime-types.ts:393`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`acp`](../packages/acp/acp), [`deepseek-account`](../packages/credentials/deepseek-account), [`goal-round-driver`](../packages/goal/goal-round-driver), [`headless`](../packages/bundle/headless), `server`, `session-controller`, [`session-telemetry`](../packages/session/session-telemetry), [`subagent`](../packages/subagent/subagent) | | `agent/inbox/claimed` | `emit` | [`packages/core/agent/src/runtime-types.ts:299`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`acp`](../packages/acp/acp), [`goal-round-driver`](../packages/goal/goal-round-driver), [`subagent`](../packages/subagent/subagent), [`tool-jobs`](../packages/jobs/tool-jobs), [`user-questions`](../packages/interaction/user-questions) | | `agent/inbox/discarded` | `emit` | [`packages/core/agent/src/runtime-types.ts:307`](../packages/core/agent/src/runtime-types.ts) | [`agent-loop`](../packages/core/agent-loop) (`emit`) | [`goal-round-driver`](../packages/goal/goal-round-driver), [`subagent`](../packages/subagent/subagent), [`user-questions`](../packages/interaction/user-questions) | diff --git a/docs/persistence-catalog.md b/docs/persistence-catalog.md index 7de9efa156..0d3b9ea113 100644 --- a/docs/persistence-catalog.md +++ b/docs/persistence-catalog.md @@ -928,9 +928,9 @@ Source: [`packages/subagent/subagent/src/descriptor.ts:38`](../packages/subagent ```ts persistence-catalog /** - * Records that this session's delegation tool exposes child provider, - * model, and reasoning-effort selection. Appended before the first model - * request; absence means the fixed-route definition. Log-only: it carries + * Records this Session’s authorized child provider/model routes. Appended + * before the first model request; absence restricts explicit selection + * to the current parent route. Log-only: it carries * no `surfaceOp` and never enters model history. */ 'subagent/model-selection-policy': { diff --git a/docs/persistence-catalog.zh.md b/docs/persistence-catalog.zh.md index 3d75f6ccfb..f48ef72c71 100644 --- a/docs/persistence-catalog.zh.md +++ b/docs/persistence-catalog.zh.md @@ -930,9 +930,9 @@ export type SessionEvent = { ```ts persistence-catalog /** - * Records that this session's delegation tool exposes child provider, - * model, and reasoning-effort selection. Appended before the first model - * request; absence means the fixed-route definition. Log-only: it carries + * Records this Session’s authorized child provider/model routes. Appended + * before the first model request; absence restricts explicit selection + * to the current parent route. Log-only: it carries * no `surfaceOp` and never enters model history. */ 'subagent/model-selection-policy': { diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 2ab2f56624..c6d4e1c326 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: e18a3e3c916afb23 - zh: 4284b50a5e9bd12f + en: 6f5f5f987c218948 + zh: 018ff71df684009c /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -225,11 +225,11 @@ en: 841abe0dd5913ea7 zh: 841abe0dd5913ea7 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/list-subagent-models: - en: c89084b31ffccbce - zh: d7b3b7000f2d2be9 + en: aba834ab4f70005a + zh: 142f819e8da5acc7 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: 9270966e03a2a62b - zh: 0544ac614aed3144 + en: c272328d36a5afaf + zh: fcf69fd0efd55ea9 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 8aa9fb0096..6b075043d7 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion available in an enabled Session. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` remains fixed-route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` uses that same policy, and without an enabled policy explicit selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2017,7 +2017,7 @@ The five read-only tools hide provider cursors and authorize every result from t ### `list_subagent_models` -Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its advertised models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. +Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. ```json { @@ -2029,7 +2029,7 @@ Discover LLM routes for subagents without changing the current Agent. Call with }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's advertised models." + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." } } } @@ -2039,7 +2039,7 @@ Source: [`packages/subagent/tool-subagent/src/list-models.ts`](../packages/subag ### `subagent` -Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. +Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. ```json { @@ -2056,6 +2056,18 @@ Delegate a self-contained task to a subagent (a separate agent that works in its "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -2067,7 +2079,7 @@ Delegate a self-contained task to a subagent (a separate agent that works in its Source: [`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion available in an enabled Session. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` remains fixed-route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. +The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` uses that same policy, and without an enabled policy explicit selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index b505568dde..c67ea4ecdf 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 则展示为已启用 Session 中可用的固定配套工具。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 始终使用固定路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 使用同一策略，没有启用策略时，显式选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2023,7 +2023,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， ### `list_subagent_models` -发现 subagent 可用的 LLM 路由，不更改当前 Agent。无参数调用会列出已注册提供方；提供 `provider` 时会列出其公布的模型；同时提供 `provider` 和 `model` 时会检查该精确模型及其推理强度。目录条目只提供建议：adapter 可能接受未列出的模型 id。把返回的 id 用于委派工具的 `provider`、`model` 与 `reasoning_effort` 字段。 +发现 subagent 可用的 LLM 路由，不更改当前 Agent。无参数调用会列出已注册提供方；提供 `provider` 时会列出其允许的模型；同时提供 `provider` 和 `model` 时会检查该精确模型及其推理强度。目录条目只提供建议：adapter 可能接受未列出的模型 id。把返回的 id 用于委派工具的 `provider`、`model` 与 `reasoning_effort` 字段。 ```json { @@ -2035,7 +2035,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's advertised models." + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." } } } @@ -2045,7 +2045,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， ### `subagent` -将一项自包含任务委派给 subagent（在自身上下文中工作的独立 agent），用它卸载聚焦且独立的工作，例如研究、限定范围的实现或分析，以免消耗当前对话的上下文。subagent 会返回结果，但不会返回中间步骤。请提供完整、独立的提示词，因为它看不到当前对话。此工具启动独立管理的 subagent 并立即返回其 id。运行时会在任务结束时通知你。子代理通过 `send_message` 汇报结果；你可以使用 `send_message` 在运行时引导它，或在结束后继续对话。 +将一项自包含任务委派给 subagent（在自身上下文中工作的独立 agent），用它卸载聚焦且独立的工作，例如研究、限定范围的实现或分析，以免消耗当前对话的上下文。subagent 会返回结果，但不会返回中间步骤。此工具启动独立管理的 subagent 并立即返回其 id。运行时会在任务结束时通知你。子代理通过 `send_message` 汇报结果；你可以使用 `send_message` 在运行时引导它，或在结束后继续对话。子级 LLM 选择是可选的。省略 `provider`、`model` 和 `reasoning_effort` 会使用配置的子级默认值，并从父 Agent 继承缺失且兼容的值。先通过 `list_subagent_models` 查询路由和强度，再同时提供 `provider` 与 `model`。更改生效路由却不指定强度时，使用所选模型的默认强度。 ```json { @@ -2062,6 +2062,18 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -2073,7 +2085,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， 来源：[`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 则展示为已启用 Session 中可用的固定配套工具。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 始终使用固定路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 +注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 使用同一策略，没有启用策略时，显式选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 diff --git a/packages/bundle/base/cordis.patch.yml b/packages/bundle/base/cordis.patch.yml index 5219c2444f..fb913cd92f 100644 --- a/packages/bundle/base/cordis.patch.yml +++ b/packages/bundle/base/cordis.patch.yml @@ -369,6 +369,9 @@ - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' + - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' @@ -378,10 +381,9 @@ provider: spawn toolName: subagent - # Fork omits model selection so provider/model stay equal to the parent and - # the inherited history remains eligible for KV Cache reuse. Parent and child + # Fork inherits the parent route by default for KV Cache reuse. Parent and child # share the messaging tools; return guidance follows the inherited history. - # See .agents/notes/implemented/feature/2026-08-18-model-selected-subagent-routes.md + # See packages/subagent/tool-subagent/README.md # and .agents/notes/implemented/architecture/2026-08-10-fork-children-stay-one-shot.md. - id: tool-subagent-fork name: '@deepseek-ai/dsh-tool-subagent' diff --git a/packages/bundle/web-app/cordis.patch.yml b/packages/bundle/web-app/cordis.patch.yml index ee5201b9c7..ec3815178e 100644 --- a/packages/bundle/web-app/cordis.patch.yml +++ b/packages/bundle/web-app/cordis.patch.yml @@ -542,6 +542,9 @@ - id: tool-subagent-control disabled: true +- id: tool-subagent-list-models + disabled: true + - id: tool-subagent-list-agents disabled: true diff --git a/packages/bundle/web-app/presets/cordis.patch.yml b/packages/bundle/web-app/presets/cordis.patch.yml index 277176f1a0..25b7ddfa0f 100644 --- a/packages/bundle/web-app/presets/cordis.patch.yml +++ b/packages/bundle/web-app/presets/cordis.patch.yml @@ -87,6 +87,8 @@ config: - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' - id: tool-subagent diff --git a/packages/bundle/web-app/presets/ptc.patch.yml b/packages/bundle/web-app/presets/ptc.patch.yml index 1aed4cb9e9..22e37cbbee 100644 --- a/packages/bundle/web-app/presets/ptc.patch.yml +++ b/packages/bundle/web-app/presets/ptc.patch.yml @@ -88,6 +88,8 @@ config: - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' - id: tool-subagent diff --git a/packages/bundle/web-app/presets/standard.patch.yml b/packages/bundle/web-app/presets/standard.patch.yml index f2a9e080fe..daf8f63591 100644 --- a/packages/bundle/web-app/presets/standard.patch.yml +++ b/packages/bundle/web-app/presets/standard.patch.yml @@ -88,6 +88,8 @@ config: - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' - id: tool-subagent diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index e390d7ddca..41e61ad7ac 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -24,14 +24,14 @@ en: 42673064ba88a211 zh: 8fa26285bd649dc9 /deepseek-ai-dsh-tool-subagent/use-this-package/selecting-a-child-llm: - en: 0d37e3b3957515e4 - zh: ea1a94b2f29e7d9a + en: b4ba9650c228b9b6 + zh: 027833db9aafd752 /deepseek-ai-dsh-tool-subagent/understand-the-implementation: en: 6ce5f749256a7e6c zh: a4f2adb04a70d9cf /deepseek-ai-dsh-tool-subagent/understand-the-implementation/design-concept: - en: 5ceacfed4ea0c648 - zh: 005284534923d854 + en: 4847e2bcdde8e282 + zh: cc833748d954a65e /deepseek-ai-dsh-tool-subagent/understand-the-implementation/activation-ownership: en: ef692c4d17290920 zh: b1d0d3dab011e2cd @@ -54,11 +54,11 @@ en: e4c7376d4f29dbd5 zh: 6b7a9eaf69c9ea53 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/what-the-model-sees: - en: 12c872e4264bb6c0 - zh: 20f870a78ebf8d29 + en: c6da697a3965147c + zh: 4f4a48054bcbda88 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/token-effect: - en: 02da2a5e0a75fd91 - zh: fd2d8b11386fc83d + en: 6a1a6351a3a88424 + zh: c574e9ee286518ed /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/kv-cache-effect: en: 46f736c6c389052a zh: a7765541030e9104 @@ -66,11 +66,11 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: 1c9dd328eaf96683 - zh: 3be88d3b00f51cfa + en: 14c12f23474ee1bf + zh: bd803212b3b0ba82 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: - en: 8137e72f630f0190 - zh: 73b1d9374b6cbccd + en: cfdfe3b1106c3b57 + zh: 3f421f2901d6f7ab /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/kv-cache-effect: en: 7dc47a34d5574fbd zh: bd7bd2a4da2f8ae2 @@ -114,8 +114,8 @@ en: f5d5f97a153f7c70 zh: 3d6bc05b9eadfbc2 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work: - en: 40e99fb684ccffc4 - zh: 08c6820c7a5343e8 + en: 93d06d9646fe848d + zh: bc86f8bd38b4b61a /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work/dev-note: en: e67f633dc6fe7773 zh: 3bb0c20c93f16f98 diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index 09fc202ba6..e407e5f7ae 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -36,6 +36,7 @@ Load the subagent service, an in-process or remote backend, and this tool; then ```yaml - name: '@deepseek-ai/dsh-subagent' - name: '@deepseek-ai/dsh-subagent-spawn-in-process' +- name: '@deepseek-ai/dsh-tool-subagent/list-models' - name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn @@ -64,7 +65,7 @@ Every backend’s completion notice carries its final answer. Local Spawn and Fo ### Selecting a child LLM -Set `modelSelectionSettings: true` to sample the Host's `subagent-model-selection` preference when each fresh top-level Session is composed. A restored Session without a recorded policy remains disabled, including an explicitly empty restore. When enabled, the non-empty exact provider/model route list is recorded in the Session, inherited by child Sessions, and unchanged by later settings edits. The tool then exposes optional `provider`, `model`, and `reasoning_effort` fields and registers the shared `list_subagent_models` tool. This mode requires a backend that advertises `agentOptions`; both in-process backends and DSH SDK support it, while ACP, Codex, and Claude Code reject it rather than ignore it. +Backends that support `agentOptions` always expose optional `provider`, `model`, and `reasoning_effort` fields. Load `@deepseek-ai/dsh-tool-subagent/list-models` once alongside the delegation tools for shared discovery; shipped compositions include it. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. Set `modelSelectionSettings: true` to sample the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. A restored Session without a policy keeps the parent-route restriction. ACP, Codex, and Claude Code do not support `agentOptions` and reject selection fields. A call supplies `provider` and `model` together, or supplies only an effort when configured, parent, or provider-owned defaults provide the route. Static `provider.agentRouteDefaults`, when present, form the provider/model baseline; tool configuration and model fields overlay it before route-aware effort merging and exact-route preflight. Providers without these defaults use compatible values from the parent's latest logged request, then the parent's creation options before its first request, while retaining the configured `maxTokens`. Changing the route without an explicit effort clears the inherited route-owned effort, so the selected model resolves its default. The live LLM adapter validates the effective route before child creation. Catalog membership remains advisory, so a model can use an unlisted id when its adapter accepts it. @@ -80,7 +81,7 @@ This section explains how the tool mirrors provider lifecycle and settles runs; ### Design concept -One instance is one provider plus one tool name. The plugin mirrors provider lifecycle: it registers the tool when the named provider appears and disposes it when the provider leaves, so sibling load order and HMR replacement cannot strand a dangling tool. Direct Agent setup passes its unpublished Session explicitly and awaits installation before publication. A settings-backed standing preset receives each matching Agent through `agent/created`, selects policy from its Session, and awaits installation through its Context; installation failure rejects creation. A numeric `maxDepth` or configured LLM selection the provider cannot enforce fails the mount instead of the first delegation. At most one instance in a tool scope may own model selection because `list_subagent_models` has a global name. +One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Settings-backed instances capture Session authorization during Agent creation; calls read the durable policy independently of registration. Direct Agent setup supplies its unpublished Session explicitly. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. Discovery has one separate composition entry shared by spawn and fork tools. No runtime invariant companion is published: policy comes directly from the Session projection and is checked at execution, with no separately retained per-Agent definition state to reconcile. ### Activation ownership @@ -128,11 +129,11 @@ Read these pages when the package-level contract is not enough; they move from t #### What the model sees -The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. An enabled Session policy adds `provider`, `model`, and `reasoning_effort` plus route guidance; the backend must support `agentOptions`. Context inheritance and continuation support determine the task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. +The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. Backends supporting `agentOptions` always expose `provider`, `model`, and `reasoning_effort`. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. #### Token effect -Fixed schema cost per parent request; model selection adds three parameters. Each tool instance adds one schema and one short system-prompt section. +Fixed schema cost per parent request, including three optional route fields on capable backends. Each tool instance adds one schema; visible delegation tools share one short system-prompt section. #### KV Cache effect @@ -142,11 +143,11 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -A settings-controlled instance whose Session carries a policy exposes the child LLM selection fields and `list_subagent_models`. Calls reject while the optional `ctx.llm` service is unavailable. Discovery returns only registered providers and advertised models in the exact route policy; an unauthorized provider is rejected before its adapter catalog is called, and an exact lookup must be allowed before it resolves the model's reasoning efforts and default. Execution independently enforces the same policy. +The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. Discovery resolves only allowed routes, including exact model ids absent from advisory catalogs; unauthorized provider/model lookups fail before adapter access. Execution independently enforces the same policy. #### Token effect -One fixed discovery schema is present in enabled compositions. Directory contents enter the transcript only when the model calls the tool. +One fixed discovery schema is present in shipped compositions, regardless of the model-selection preference. Route metadata enters the transcript only when the model calls the tool. #### KV Cache effect @@ -208,8 +209,8 @@ Append-only; newly visible content follows the reusable request prefix and does These limits define what this tool does not return or enforce; they are current package constraints. - **This tool does not wait for results** — use the completion notice or child messages. Programmatic result waiting belongs to the subagent service. -- **Shipped fork tools cannot select a child LLM route** — they inherit the parent's provider and model to keep the copied conversation prefix eligible for KV Cache reuse. Re-enable selection only when route changes preserve reuse or expose a bounded recomputation cost. -- **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. LLM selection requires an enabled per-Session preference and a provider that advertises `agentOptions`; both in-process providers and DSH SDK advertise it, while ACP, Codex, and Claude Code reject it rather than ignore it. +- **Fork route changes can lose KV Cache reuse** — optional route fields follow the same Session policy as spawn; changing provider or model may require prefilling the inherited conversation again. +- **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection is restricted to the Session’s authorized routes, or the current parent route without an enabled policy; unsupported backends reject selection fields. ### Dev Note diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index 4b695ccf37..ea0a06a3ba 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -36,6 +36,7 @@ kind: "package-reference" ```yaml - name: '@deepseek-ai/dsh-subagent' - name: '@deepseek-ai/dsh-subagent-spawn-in-process' +- name: '@deepseek-ai/dsh-tool-subagent/list-models' - name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn @@ -64,7 +65,7 @@ kind: "package-reference" ### 选择子级 LLM -设置 `modelSelectionSettings: true`，即可在组合每个全新顶层 Session 时读取宿主的 `subagent-model-selection` 偏好。没有已记录策略的恢复 Session 会保持禁用，包括显式为空的恢复。启用后，非空的精确 provider/model 路由列表会记录进 Session、由子 Session 继承，后续设置编辑不会改变它。工具随后公开可选的 `provider`、`model` 与 `reasoning_effort` 字段，并注册共享的 `list_subagent_models` 工具。此模式要求后端声明 `agentOptions`；两个进程内后端和 DSH SDK 支持该能力，而 ACP、Codex 与 Claude Code 会拒绝它，而不是忽略它。 +支持 `agentOptions` 的后端始终公开可选的 `provider`、`model` 和 `reasoning_effort` 字段。在委派工具旁加载一次 `@deepseek-ai/dsh-tool-subagent/list-models` 即可共享发现功能；随附组合已包含它。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。设置 `modelSelectionSettings: true` 后，新顶层 Session 会读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有策略的恢复 Session 仍受父级路由限制。ACP、Codex 和 Claude Code 不支持 `agentOptions`，会拒绝选择字段。 一次调用需同时提供 `provider` 与 `model`；当配置值、父 agent 值或提供方持有的默认值能提供路由时，也可只提供推理等级。静态的 `provider.agentRouteDefaults` 在存在时构成提供方／模型基线；工具配置与模型字段会在路由相关强度合并和确切路由预检前覆盖它。没有这些默认值的提供方会使用父 agent 最新已记录请求中的兼容值，再使用父级首次请求前的创建选项，并保留配置的 `maxTokens`。更改路由但未显式提供推理等级时，会清除继承的路由自有等级，使所选模型解析自己的默认值。实时 LLM 适配器在创建子 agent 前校验有效路由。目录成员资格只提供建议，因此适配器接受时，模型可以使用未列出的 id。 @@ -80,7 +81,7 @@ kind: "package-reference" ### 设计理念 -一个实例就是一个提供方加一个工具名称。插件镜像提供方生命周期：具名提供方出现时注册工具，提供方离开时释放工具，因此同级加载顺序与 HMR 替换不会让工具悬空。直接 Agent setup 显式传入尚未发布的 Session，并在发布前等待安装完成。由设置控制的常驻 preset 通过 `agent/created` 接收每个匹配 Agent，从其 Session 选择策略，并等待通过其 Context 发起的安装；安装失败会拒绝创建。提供方无法执行的数值型 `maxDepth` 或已配置 LLM 选择会在挂载时失败，而不是在首次委派时失败。每个工具作用域内最多一个实例可以拥有模型选择，因为 `list_subagent_models` 使用全局名称。 +一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。读取设置的实例在 Agent 创建时记录 Session 授权；调用时独立于注册读取持久策略。直接配置 Agent 时显式传入尚未发布的 Session。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。发现功能由 spawn 和 fork 工具共享的独立组合项提供。本包不导出运行时不变量：策略直接来自 Session 投影，并在执行时校验，没有另行保存的逐 Agent 定义状态需要协调。 ### Activation 所有权 @@ -128,11 +129,11 @@ kind: "package-reference" #### 模型看到什么 -提供方存在时，以配置的工具名称公开生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent)。启用的 Session 策略会添加 `provider`、`model`、`reasoning_effort` 和路由指引；后端必须支持 `agentOptions`。上下文继承和继续执行能力决定任务与后续消息的描述。工具限制会同时移除 schema 及其 `tool:` 指引 section。 +生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。支持 `agentOptions` 的后端始终公开 `provider`、`model` 和 `reasoning_effort`。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 #### Token 影响 -每个父级请求支付固定的 schema 成本；模型选择增加三个参数。每个工具实例增加一个 schema 和一个简短的系统提示词 section。 +每次父级请求承担固定 schema 成本，其中支持选择的后端包含三个可选路由字段。每个工具实例增加一份 schema；可见委派工具共享一个简短系统提示词段。 #### KV Cache 影响 @@ -142,11 +143,11 @@ kind: "package-reference" #### 模型看到什么 -Session 携带策略的 settings 控制实例会公开子级 LLM 选择字段与 `list_subagent_models`。可选 `ctx.llm` 服务不可用时，调用会失败。发现只返回精确路由策略中的已注册提供方与已公布模型；未授权提供方会在调用其适配器目录前被拒绝，精确查询也必须先获准，才会解析模型的推理强度与默认值。执行阶段会独立强制同一策略。 +共享的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。发现只解析允许的路由，包括建议性目录未列出的精确模型 ID；未授权的提供方／模型查询会在访问适配器前失败。执行过程独立强制同一策略。 #### Token 影响 -启用的组合中存在一个固定发现 schema。只有模型调用工具时，目录内容才进入 transcript。 +随附组合中始终存在一份固定发现 schema，不受模型选择偏好影响。只有模型调用工具时，路由元数据才进入 transcript。 #### KV Cache 影响 @@ -208,8 +209,8 @@ Start independent delegations with `subagent` or `subagent_fork` together in one 这些限制说明本工具不返回或不强制执行什么；它们是当前包约束。 - **本工具不等待结果**——使用完成通知或子任务消息。程序化结果等待由 subagent 服务提供。 -- **随附 fork 工具不能选择子级 LLM 路由**——它们继承父级提供方与模型，使复制的对话前缀仍有资格复用 KV Cache。仅当路由变更能保留复用或公开有界重算成本时，才重新启用选择。 -- **非路由子 agent 策略按实例固定**——另一个 persona、工具过滤器或深度上限需要另一个名称不同的工具。LLM 选择要求启用逐 Session 偏好，且提供方必须声明 `agentOptions`；两个进程内提供方和 DSH SDK 会声明该能力，而 ACP、Codex 与 Claude Code 会拒绝它，而不是忽略它。 +- **更改 fork 路由可能失去 KV Cache 复用** — 可选路由字段遵循与 spawn 相同的 Session 策略；更改提供方或模型可能需要重新预填充继承的对话。 +- **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择受 Session 授权路由限制，没有启用策略时只允许当前父级路由；不支持选择的后端会拒绝选择字段。 ### 开发备注 diff --git a/packages/subagent/tool-subagent/package.json b/packages/subagent/tool-subagent/package.json index f6c8100036..0117a36a9a 100644 --- a/packages/subagent/tool-subagent/package.json +++ b/packages/subagent/tool-subagent/package.json @@ -23,10 +23,15 @@ "default": "./lib/model-selection-settings.js" }, "./src/*": "./src/*", - "./package.json": "./package.json" + "./package.json": "./package.json", + "./list-models": { + "types": "./lib/types/list-models.d.ts", + "default": "./lib/list-models.js" + } }, "files": [ "lib/index.js", + "lib/list-models.js", "lib/model-selection-settings.js", "lib/types/**/*.d.ts" ], diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index 968b68ebf3..90f98d10aa 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -11,7 +11,7 @@ import z from '@deepseek-ai/schemastery' import { scopeChainOf, scopeOf } from '@deepseek-ai/dsh-scope' import { defineTool } from '@deepseek-ai/dsh-tools' import type { ToolDefinition } from '@deepseek-ai/dsh-tools' -import type { Agent, AgentOptions } from '@deepseek-ai/dsh-agent' +import type { AgentOptions } from '@deepseek-ai/dsh-agent' import { ReasoningEffortId } from '@deepseek-ai/dsh-llm' import type { ContentBlock } from '@deepseek-ai/dsh-llm' import { SessionSeq } from '@deepseek-ai/dsh-session' @@ -28,8 +28,7 @@ import { preflightChildLlmRoute, requestedAgentOptions, } from './model-selection.ts' -import type { DelegationModelRequest, ModelSelectionPolicy } from './model-selection.ts' -import { registerListSubagentModels } from './list-models.ts' +import type { DelegationModelRequest } from './model-selection.ts' import type {} from './model-selection-settings.ts' import { recordSubagentModelSelection, @@ -201,14 +200,13 @@ export function apply(ctx: Context, config: Config, session?: Session): void { const initialProvider = ctx.subagents.getProvider(config.provider) if (initialProvider !== undefined) assertSubagentProviderConfiguration(initialProvider) - const install = (runtimeCtx: Context, modelSelectionPolicy: ModelSelectionPolicy | undefined): void => { - const modelSelectionEnabled = modelSelectionPolicy !== undefined - if (modelSelectionPolicy !== undefined) registerListSubagentModels(runtimeCtx, modelSelectionPolicy) + const install = (runtimeCtx: Context): void => { // Load order and HMR replacement can change provider availability while // this fiber remains active. let mounted: { subagentProvider: SubagentProvider; disposeTool: () => void } | undefined const mount = (subagentProvider: SubagentProvider): void => { assertSubagentProviderConfiguration(subagentProvider) + const modelSelectionEnabled = subagentProvider.capabilities.agentOptions const wording = providerWording(subagentProvider.inheritsParentContext) const providerRouteDefaults = subagentProvider.agentRouteDefaults const selectionDescription = providerRouteDefaults !== undefined @@ -284,6 +282,8 @@ export function apply(ctx: Context, config: Config, session?: Session): void { throw new Error('subagent tool requires a calling agent (exec.agent was undefined)') } + const routes = subagentModelSelectionPolicy(runtimeCtx.sessionProjections, parent.session) + const modelSelectionPolicy = routes === undefined ? undefined : { routes } const modelRequest = args as DelegationModelRequest const parentOptions = parentAgentOptionsForDelegation(parent) const requiresRoutePreflight = hasDelegationModelRequest(modelRequest) @@ -386,7 +386,7 @@ export function apply(ctx: Context, config: Config, session?: Session): void { } if (config.modelSelectionSettings !== true) { - install(ctx, undefined) + install(ctx) return } @@ -397,7 +397,7 @@ export function apply(ctx: Context, config: Config, session?: Session): void { + '@deepseek-ai/dsh-tool-subagent/model-selection-settings in the Host scope', ) } - const selectForSession = (target: Session): ModelSelectionPolicy | undefined => { + const selectForSession = (target: Session): void => { const freshSession = target.firstLiveSeq === 0 // oxlint-disable-next-line typescript/no-deprecated -- Existing Session history read; migration deferred. && target.eventAt(SessionSeq(0))?.type !== 'session/end-seed' @@ -423,11 +423,11 @@ export function apply(ctx: Context, config: Config, session?: Session): void { if (allowedModels !== undefined) { recordSubagentModelSelection(ctx.sessionProjections, target, allowedModels) } - return allowedModels === undefined ? undefined : { routes: allowedModels } } if (session !== undefined) { - install(ctx, selectForSession(session)) + selectForSession(session) + install(ctx) return } @@ -438,54 +438,21 @@ export function apply(ctx: Context, config: Config, session?: Session): void { const agents = ctx.get('agents') /* v8 ignore next -- shipped preset compositions always include the Agent registry. */ if (agents === undefined) throw new Error('tool-subagent: standing `modelSelectionSettings` requires the Agent registry') - const scopedInstalls = new WeakMap>() - const installing = new WeakSet() - const belongsToComposition = (candidate: Agent): boolean => - scopeChainOf(scopeOf(candidate.ctx)).includes(compositionScope) - const installScoped = (candidate: Agent): ReturnType | undefined => { - const existing = scopedInstalls.get(candidate) - if (existing !== undefined) return existing - if (installing.has(candidate)) return - // Reserve before the injected fiber runs: tool registration emits - // `tools/change` synchronously, which re-enters the reconciliation below. - installing.add(candidate) - let fiber: ReturnType - try { - const policy = selectForSession(candidate.session) - fiber = candidate.ctx.inject(['tools', 'subagents', 'systemPrompt'], (runtimeCtx) => { - install(runtimeCtx, policy) - }) - } finally { - installing.delete(candidate) - } - scopedInstalls.set(candidate, fiber) - return fiber + const captured = new WeakSet() + const capture = (target: Session): void => { + if (captured.has(target)) return + selectForSession(target) + captured.add(target) } - const removeScoped = (candidate: Agent): void => { - const fiber = scopedInstalls.get(candidate) - if (fiber === undefined) return - scopedInstalls.delete(candidate) - /* v8 ignore next 3 -- Cordis Fiber disposal contains registration cleanup failures; this is the final diagnostic sink. */ - void fiber.dispose().catch((error: unknown) => { - ctx.logger.warn(`tool-subagent: failed to remove recomposed Agent "${candidate.id}" definitions: ${String(error)}`) - }) - } - const reconcileComposedAgents = (): void => { - for (const candidate of agents.list()) { - if (belongsToComposition(candidate)) installScoped(candidate) - else removeScoped(candidate) + const captureComposedSessions = (): void => { + for (const agent of agents.list()) { + if (scopeChainOf(scopeOf(agent.ctx)).includes(compositionScope)) capture(agent.session) } } - // The preset-scoped listener admits descendant Agents and installs the - // sampled tool definition in each Agent's own scope, so a later settings - // change cannot mutate a live session. - ctx.on('agent/created', async ({ agent: created }) => { - await installScoped(created) - }) - ctx.on('agent/disposed', ({ agent: disposed }) => { removeScoped(disposed) }) - // Reparenting an Agent between standing presets changes its inherited tool - // set and emits `tools/change`; reconcile the Agent-owned override with the - // new ancestry. Other registry changes are idempotent no-ops here. - ctx.on('tools/change', reconcileComposedAgents) - reconcileComposedAgents() + ctx.on('agent/created', ({ agent }) => { capture(agent.session) }) + // Blank Sessions can join a preset after creation; absence also records the + // sampled off decision for this plugin lifetime. + ctx.on('tools/change', captureComposedSessions) + captureComposedSessions() + install(ctx) } diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index a1265934d3..c642c4a743 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -4,6 +4,9 @@ import type { Context } from '@deepseek-ai/cordis' import type LlmRuntime from '@deepseek-ai/dsh-llm' import type { LlmProviderInfo } from '@deepseek-ai/dsh-llm' import { defineTool } from '@deepseek-ai/dsh-tools' +import { parentAgentOptionsForDelegation } from '@deepseek-ai/dsh-subagent' +import type { ToolRunContext } from '@deepseek-ai/dsh-tools' +import { subagentModelSelectionPolicy, subagentModelSelectionProjectionDefinition } from './model-selection-state.ts' import type { ModelSelectionPolicy } from './model-selection.ts' interface ListSubagentModelsRequest { @@ -32,7 +35,7 @@ function modelLine(provider: string, model: { id: string; name: string; descript return `${provider}/${model.id} — ${model.name}${model.description === undefined ? '' : `: ${model.description}`}` } -/** Read the requested provider, advertised models, or exact-model efforts. */ +/** Read the requested provider, allowed models, or exact-model efforts. */ async function listSubagentModels( ctx: Context, policy: ModelSelectionPolicy, @@ -60,11 +63,8 @@ async function listSubagentModels( } const provider = registeredProvider(llm, policy, request.provider) if (request.model === undefined) { - const models = (await llm.listModels(provider.id)) - .filter(model => allowedRoutes.some(route => route.model === model.id)) - return models.length === 0 - ? `(no advertised models for ${provider.id})` - : models.map(model => modelLine(provider.id, model)).join('\n') + const models = await Promise.all(allowedRoutes.map(route => llm.resolveModelInfo(provider.id, route.model, signal))) + return models.map(model => modelLine(provider.id, model)).join('\n') } if (request.model.length === 0) throw new Error('`model` must be non-empty') if (!allowedRoutes.some(route => route.model === request.model)) { @@ -79,16 +79,16 @@ async function listSubagentModels( } /** - * Register `list_subagent_models` for one owning delegation-tool instance. + * Register `list_subagent_models` in the owning composition. * @param ctx - Context whose tool registry owns the fixed discovery definition. - * @param policy - Route policy captured for this Session. + * @param policyFor - Resolve the calling Session’s authorized routes. */ -export function registerListSubagentModels(ctx: Context, policy: ModelSelectionPolicy): void { +export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolRunContext) => ModelSelectionPolicy): void { ctx.tools.register(defineTool({ name: 'list_subagent_models', description: 'Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list ' - + 'registered providers, with `provider` to list its advertised models, or with `provider` and `model` ' + + 'registered providers, with `provider` to list its allowed models, or with `provider` and `model` ' + 'to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may ' + 'accept an unlisted model id. Use the returned ids with a delegation tool\'s `provider`, `model`, and ' + '`reasoning_effort` fields.', @@ -99,7 +99,7 @@ export function registerListSubagentModels(ctx: Context, policy: ModelSelectionP }, model: { type: 'string', - description: 'Exact model id to inspect. Requires provider; omit to list that provider\'s advertised models.', + description: 'Exact model id to inspect. Requires provider; omit to list that provider\'s allowed models.', }, }, output: { @@ -107,7 +107,30 @@ export function registerListSubagentModels(ctx: Context, policy: ModelSelectionP render: (_args, result) => [{ type: 'text', text: result }], }, execute(args, exec) { - return listSubagentModels(ctx, policy, args, exec.signal) + return listSubagentModels(ctx, policyFor(exec), args, exec.signal) }, })) } + +/** Shared discovery plugin for all delegation tools in one composition. */ +export const name = 'tool-subagent-list-models' +/** Services used to register discovery and read durable route policy. */ +export const inject = ['tools', 'sessionProjections'] + +/** + * Register one discovery tool whose results follow the calling Session’s policy. + * @param ctx - Owning composition context. + */ +export function apply(ctx: Context): void { + ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) + registerListSubagentModels(ctx, (exec) => { + if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') + const routes = subagentModelSelectionPolicy(ctx.sessionProjections, exec.agent.session) + if (routes !== undefined) return { routes } + const { provider, model } = parentAgentOptionsForDelegation(exec.agent) + if (provider === undefined || model === undefined) { + throw new Error('cannot discover child LLM routes without a parent provider and model') + } + return { routes: [{ provider, model }] } + }) +} diff --git a/packages/subagent/tool-subagent/src/model-selection-state.ts b/packages/subagent/tool-subagent/src/model-selection-state.ts index d3b04b828b..0854425c83 100644 --- a/packages/subagent/tool-subagent/src/model-selection-state.ts +++ b/packages/subagent/tool-subagent/src/model-selection-state.ts @@ -9,9 +9,9 @@ import { assertAllowedModelRoutes, type AllowedModelRoute } from './model-select declare module '@deepseek-ai/dsh-session/types' { interface SessionEventMap { /** - * Records that this session's delegation tool exposes child provider, - * model, and reasoning-effort selection. Appended before the first model - * request; absence means the fixed-route definition. Log-only: it carries + * Records this Session’s authorized child provider/model routes. Appended + * before the first model request; absence restricts explicit selection + * to the current parent route. Log-only: it carries * no `surfaceOp` and never enters model history. */ 'subagent/model-selection-policy': { @@ -23,7 +23,7 @@ declare module '@deepseek-ai/dsh-session/types' { declare module '@deepseek-ai/dsh-session-projection/types' { interface SessionProjectionStateMap { - /** Exact routes authorized for child LLM selection, or null when disabled. */ + /** Exact routes authorized for child LLM selection, or null for the parent route. */ subagentModelSelectionPolicy: AllowedModelRoute[] | null } } @@ -51,10 +51,10 @@ export const subagentModelSelectionProjectionDefinition = { } satisfies ProjectionDefinition<'subagentModelSelectionPolicy', AllowedModelRoute[] | null> /** - * Read the exact route list captured for a model-selectable definition. + * Read the exact route list authorized for a Session. * @param projections - registry that owns the policy projection. * @param session - session whose durable decision is read. - * @returns a detached route list, or undefined for the fixed-route definition. + * @returns a detached route list, or undefined to restrict selection to the parent route. */ export function subagentModelSelectionPolicy( projections: Pick, @@ -64,10 +64,10 @@ export function subagentModelSelectionPolicy( } /** - * Append the route policy once, before its definition can reach a model request. + * Append the route policy once, before the Session’s first model request. * @param projections - registry that owns the policy projection. - * @param session - session receiving the model-selectable definition. - * @param allowedModels - exact routes the definition may select explicitly. + * @param session - session receiving route authorization. + * @param allowedModels - exact routes the Session may select explicitly. */ export function recordSubagentModelSelection( projections: Pick, diff --git a/packages/subagent/tool-subagent/src/model-selection.ts b/packages/subagent/tool-subagent/src/model-selection.ts index 5f30b71742..04f4d33b5f 100644 --- a/packages/subagent/tool-subagent/src/model-selection.ts +++ b/packages/subagent/tool-subagent/src/model-selection.ts @@ -19,7 +19,7 @@ export const AllowedModelRouteSchema: z = z.object({ model: z.string().min(1).required(), }) -/** Route-selection authority captured by one delegation definition. */ +/** Exact routes authorized for one Session. */ export interface ModelSelectionPolicy { /** Exact provider/model routes authorized for explicit selection. */ readonly routes: readonly AllowedModelRoute[] @@ -131,7 +131,7 @@ export function requestedAgentOptions( * Enforce a settings-owned route list at the operation that creates the child. * Pure inheritance remains outside this policy because no model-facing choice * occurred; any explicit route or effort field must resolve to an allowed route. - * @param policy - Selection authority captured for this Session. + * @param policy - Session allowlist; omission permits only the current parent route. * @param parentOptions - Current parent values that supply missing child values. * @param requested - Effective child options after request/config merging. * @param request - Model-facing selection fields from the tool call. @@ -142,13 +142,16 @@ export function assertAllowedModelSelection( requested: AgentOptions | undefined, request: DelegationModelRequest, ): void { - if (policy === undefined || !hasDelegationModelRequest(request)) return + if (!hasDelegationModelRequest(request)) return const provider = requested?.provider ?? parentOptions.provider const model = requested?.model ?? parentOptions.model if (provider === undefined || model === undefined) { throw new Error('cannot select child LLM values without an effective provider and model') } - if (policy.routes.some(route => route.provider === provider && route.model === model)) return + const allowed = policy === undefined + ? provider === parentOptions.provider && model === parentOptions.model + : policy.routes.some(route => route.provider === provider && route.model === model) + if (allowed) return throw new Error(`child LLM route "${provider}/${model}" is not allowed for this Session`) } diff --git a/packages/subagent/tool-subagent/tests/harness.ts b/packages/subagent/tool-subagent/tests/harness.ts index 422d65ba0b..061161673e 100644 --- a/packages/subagent/tool-subagent/tests/harness.ts +++ b/packages/subagent/tool-subagent/tests/harness.ts @@ -9,6 +9,7 @@ import SubagentRuntime from '@deepseek-ai/dsh-subagent' import { SessionId } from '@deepseek-ai/dsh-session' import * as mock from './scripted-provider.ts' import * as tool from '../src/index.ts' +import * as discovery from '../src/list-models.ts' import SubagentModelSelectionConfig from '../src/model-selection-settings.ts' const contexts = new Set() @@ -58,6 +59,7 @@ export async function setup(toolConfig: SetupConfig, mockConfig: Partial ({ routes })) return ctx } @@ -78,13 +78,13 @@ async function setupAllowedListTool() { await ctx.plugin(LlmRuntime) await ctx.plugin(SystemPrompt) await ctx.plugin(ToolRuntime) - registerListSubagentModels(ctx, { + registerListSubagentModels(ctx, () => ({ routes: [ { provider: 'alpha', model: 'fast' }, { provider: 'alpha', model: 'unlisted' }, { provider: 'missing', model: 'hidden' }, ], - }) + })) return ctx } @@ -100,7 +100,7 @@ function call(ctx: Context, args: unknown) { } describe('list_subagent_models', () => { - it('is omitted unless its delegation-tool instance owns discovery', async () => { + it('is loaded through its own composition entry', async () => { const ctx = new Context() await ctx.plugin(SessionProjectionRegistry) await ctx.plugin(LlmRuntime) @@ -112,13 +112,43 @@ describe('list_subagent_models', () => { expect(ctx.tools.get('list_subagent_models')).toBeUndefined() }) + it('requires a calling Agent and an effective parent route', async () => { + const ctx = new Context() + await ctx.plugin(SessionProjectionRegistry) + await ctx.plugin(SystemPrompt) + await ctx.plugin(ToolRuntime) + await ctx.plugin({ apply, inject, name }) + expect(text(await call(ctx, {}))).toContain('requires a calling agent') + await ctx.fiber.dispose() + + const parentContext = await setup({ provider: 'mock' }) + const agent = modelSelectionSetupAgent(parentContext) + const result = await parentContext.tools.execute({ + signal: testToolSignal, callId: ToolCallId('missing-parent-route'), + name: 'list_subagent_models', arguments: {}, agent, + }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('without a parent provider and model') + }) + + it('reads the recorded Session allowlist through the shared discovery definition', async () => { + const ctx = await setup({ provider: 'mock', withModelSelection: true }) + ctx.llm.registerAdapter(['alpha', 'secret'], new CatalogAdapter()) + const result = await ctx.tools.execute({ + signal: testToolSignal, callId: ToolCallId('recorded-routes'), + name: 'list_subagent_models', arguments: {}, agent: modelSelectionSetupAgent(ctx), + }) + expect(result.isError).toBe(false) + expect(text(result)).toBe('alpha — ALPHA API') + }) + it('stays registered without the optional LLM service and rejects discovery calls', async () => { const ctx = new Context() await ctx.plugin(SystemPrompt) await ctx.plugin(ToolRuntime) await mountWorkingDirectoryFixture(ctx) await ctx.plugin(SubagentRuntime) - registerListSubagentModels(ctx, { routes: [{ provider: 'alpha', model: 'fast' }] }) + registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'alpha', model: 'fast' }] })) const result = await call(ctx, {}) expect(result.isError).toBe(true) expect(text(result)).toContain('`llm` service is unavailable') @@ -127,7 +157,7 @@ describe('list_subagent_models', () => { it('rejects two discovery-owning instances in one tool scope', async () => { const ctx = await setupListTool() expect(() => { - registerListSubagentModels(ctx, { routes: [{ provider: 'alpha', model: 'fast' }] }) + registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'alpha', model: 'fast' }] })) }).toThrow('tool "list_subagent_models" is already registered') }) @@ -164,7 +194,7 @@ describe('list_subagent_models', () => { ctx.llm.registerAdapter(['alpha', 'beta'], new CatalogAdapter()) expect(text(await call(ctx, {}))).toBe('alpha — ALPHA API') - expect(text(await call(ctx, { provider: 'alpha' }))).toBe('alpha/fast — Fast: Focused work.') + expect(text(await call(ctx, { provider: 'alpha' }))).toBe('alpha/fast — Fast: Focused work.\nalpha/unlisted — Fast: Focused work.') expect(text(await call(ctx, { provider: 'alpha', model: 'unlisted' }))) .toContain('alpha/unlisted — Fast') @@ -186,12 +216,12 @@ describe('list_subagent_models', () => { expect(listModels).not.toHaveBeenCalled() }) - it('renders an empty advertised model list', async () => { + it('resolves an allowed model even when the advertised catalog is empty', async () => { const ctx = await setupListTool([{ provider: 'alpha', model: 'fast' }]) ctx.llm.registerAdapter(['alpha'], new CatalogAdapter(true)) const result = await call(ctx, { provider: 'alpha' }) expect(result.isError).toBe(false) - expect(text(result)).toBe('(no advertised models for alpha)') + expect(text(result)).toBe('alpha/fast — Fast: Focused work.') }) it('inspects exact-model efforts, descriptions, and defaults', async () => { diff --git a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts index 31ab63b973..a4dc33b0fa 100644 --- a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts @@ -14,6 +14,7 @@ import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection' import SubagentRuntime from '@deepseek-ai/dsh-subagent' import * as SubagentSpawn from '@deepseek-ai/dsh-subagent-spawn-in-process' import * as tool from '../src/index.ts' +import * as discovery from '../src/list-models.ts' import SubagentModelSelectionConfig from '../src/model-selection-settings.ts' import { subagentModelSelectionPolicy, @@ -48,8 +49,10 @@ async function boot(withPreset = true): Promise { await mountWorkingDirectoryFixture(ctx) subagentConfigs.set(ctx, await liveConfig(ctx, SubagentRuntime)) await ctx.plugin(SubagentSpawn, { providerName: 'spawn' }) + await ctx.plugin(discovery) if (withPreset) { const preset = createScope(ctx, { preset: 'model-selection-test' }) + await preset.ctx.plugin(discovery) await preset.ctx.plugin(tool, { provider: 'spawn', modelSelectionSettings: true, @@ -136,7 +139,7 @@ describe('SubagentModelSelectionConfig', () => { it('samples each new root Session without changing existing definitions', async () => { const ctx = await boot() const disabled = await createAgent(ctx, 'disabled') - expect(selectable(ctx, disabled)).toBe(false) + expect(selectable(ctx, disabled)).toBe(true) expect(subagentModelSelectionPolicy(ctx.sessionProjections, disabled.session)).toBeUndefined() await selectionConfigs.get(ctx)!.update({ @@ -146,11 +149,11 @@ describe('SubagentModelSelectionConfig', () => { const enabled = await createAgent(ctx, 'enabled') expect(subagentModelSelectionPolicy(ctx.sessionProjections, enabled.session)).toEqual(ALLOWED_MODELS) expect(selectable(ctx, enabled)).toBe(true) - expect(selectable(ctx, disabled)).toBe(false) + expect(selectable(ctx, disabled)).toBe(true) await selectionConfigs.get(ctx)!.update({ enabled: false }) const disabledAgain = await createAgent(ctx, 'disabled-again') - expect(selectable(ctx, disabledAgain)).toBe(false) + expect(selectable(ctx, disabledAgain)).toBe(true) expect(selectable(ctx, enabled)).toBe(true) await ctx.fiber.dispose() }) @@ -187,7 +190,35 @@ describe('SubagentModelSelectionConfig', () => { await ctx.fiber.dispose() }) - it('installs one tool when recording the Session policy triggers a registry refresh', async () => { + it('captures a blank Session joining the preset without resampling existing Sessions', async () => { + const ctx = await boot() + const disabled = await createAgent(ctx, 'already-disabled') + const joining = await ctx.agents.create({ sessionId: SessionId('joining-preset') }) + await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: ALLOWED_MODELS }) + const preset = modelSelectionPresets.get(ctx)! + bindScopeParent(scopeOf(joining.agent.ctx)!, scopeOf(preset.ctx)!) + ctx.emit(scopeTarget({}, scopeOf(preset.ctx)), 'tools/change') + expect(subagentModelSelectionPolicy(ctx.sessionProjections, joining.agent.session)).toEqual(ALLOWED_MODELS) + expect(subagentModelSelectionPolicy(ctx.sessionProjections, disabled.session)).toBeUndefined() + await ctx.fiber.dispose() + }) + + it('captures policy for existing Agents only in the mounting preset', async () => { + const ctx = await boot(false) + const preset = createScope(ctx, { preset: 'late-tools' }) + const inside = await ctx.agents.create({ + sessionId: SessionId('existing-inside'), + setup: (agentCtx) => { bindScopeParent(scopeOf(agentCtx)!, scopeOf(preset.ctx)!) }, + }) + const outside = await ctx.agents.create({ sessionId: SessionId('existing-outside') }) + await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: ALLOWED_MODELS }) + await preset.ctx.plugin(tool, { provider: 'spawn', modelSelectionSettings: true }) + expect(subagentModelSelectionPolicy(ctx.sessionProjections, inside.agent.session)).toEqual(ALLOWED_MODELS) + expect(subagentModelSelectionPolicy(ctx.sessionProjections, outside.agent.session)).toBeUndefined() + await ctx.fiber.dispose() + }) + + it('does not register tools when recording the Session policy triggers a registry refresh', async () => { const ctx = await boot() try { await selectionConfigs.get(ctx)!.update({ @@ -205,7 +236,7 @@ describe('SubagentModelSelectionConfig', () => { const agent = await createAgent(ctx, 'policy-refresh') expect(selectable(ctx, agent)).toBe(true) - expect(register.mock.calls.filter(([definition]) => definition.name === 'subagent')).toHaveLength(1) + expect(register).not.toHaveBeenCalled() } finally { await ctx.fiber.dispose() } @@ -237,10 +268,11 @@ describe('SubagentModelSelectionConfig', () => { await ctx.fiber.dispose() }) - it('installs per-Agent definitions for a shared preset scope', async () => { + it('shares definitions across Agents and lets scope ancestry own their visibility', async () => { const ctx = await boot(false) const preset = createScope(ctx, { preset: 'standard' }) const other = createScope(ctx, { preset: 'minimal' }) + await preset.ctx.plugin(discovery) await preset.ctx.plugin(tool, { provider: 'spawn', modelSelectionSettings: true, @@ -256,14 +288,16 @@ describe('SubagentModelSelectionConfig', () => { }) const disabled = await createComposed('preset-disabled') - expect(selectable(ctx, disabled.agent)).toBe(false) + expect(selectable(ctx, disabled.agent)).toBe(true) await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: ALLOWED_MODELS, }) const enabled = await createComposed('preset-enabled') expect(selectable(ctx, enabled.agent)).toBe(true) - expect(selectable(ctx, disabled.agent)).toBe(false) + expect(selectable(ctx, disabled.agent)).toBe(true) + expect(ctx.tools.get('subagent', enabled.agent)).toBe(ctx.tools.get('subagent', disabled.agent)) + expect(ctx.tools.get('subagent', enabled.agent)).toBe(ctx.tools.get('subagent', scopeOf(preset.ctx))) enabledBinding!.rebind(scopeOf(other.ctx)!) ctx.emit(scopeTarget({}, scopeOf(preset.ctx)), 'tools/change') @@ -291,37 +325,31 @@ describe('SubagentModelSelectionConfig', () => { await ctx.fiber.dispose() }) - it('releases a shared-preset installation reservation after policy selection fails', async () => { - const ctx = await boot(false) - const preset = createScope(ctx, { preset: 'standard' }) - const other = createScope(ctx, { preset: 'minimal' }) - await preset.ctx.plugin(tool, { - provider: 'spawn', - modelSelectionSettings: true, - }) - let binding: ReturnType | undefined - const handle = await ctx.agents.create({ - sessionId: SessionId('preset-policy-retry'), - setup: (agentCtx) => { - binding = bindScopeParent(scopeOf(agentCtx)!, scopeOf(preset.ctx)!) - }, - }) - expect(selectable(ctx, handle.agent)).toBe(false) - - binding!.rebind(scopeOf(other.ctx)!) - ctx.emit(scopeTarget({}, scopeOf(preset.ctx)), 'tools/change') - binding!.rebind(scopeOf(preset.ctx)!) - vi.spyOn(ctx.subagentModelSelection, 'current') - .mockImplementationOnce(() => { throw new Error('transient settings read') }) - .mockReturnValue({ enabled: true, allowedModels: ALLOWED_MODELS }) - - expect(() => { ctx.emit(scopeTarget({}, scopeOf(preset.ctx)), 'tools/change') }) - .toThrow('transient settings read') - ctx.emit(scopeTarget({}, scopeOf(preset.ctx)), 'tools/change') - await vi.waitFor(() => { expect(selectable(ctx, handle.agent)).toBe(true) }) - - await handle.dispose() - await ctx.fiber.dispose() + it('keeps shared delegation and discovery hidden by the existing inherited-tool filter', async () => { + const ctx = await boot() + try { + const parent = await createAgent(ctx, 'filter-parent') + const preset = modelSelectionPresets.get(ctx)! + const child = (await ctx.agents.create({ + sessionId: SessionId('filter-child'), + meta: { parentSession: parent.id, origin: 'subagent' }, + setup: (childCtx) => { + bindScopeParent(scopeOf(childCtx)!, scopeOf(preset.ctx)!) + childCtx.tools.restrict({ deny: ['subagent', 'list_subagent_models'] }) + }, + })).agent + expect(ctx.tools.get('subagent', parent)).toBeDefined() + expect(ctx.tools.get('subagent', child)).toBeUndefined() + expect(ctx.tools.get('list_subagent_models', child)).toBeUndefined() + const result = await ctx.tools.execute({ + name: 'subagent', callId: ToolCallId('filtered-delegation'), arguments: {}, + agent: child, signal: new AbortController().signal, + }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('unknown tool') + } finally { + await ctx.fiber.dispose() + } }) it('inherits the parent decision and preserves seeded decisions across composition', async () => { @@ -341,7 +369,7 @@ describe('SubagentModelSelectionConfig', () => { const orphan = await createAgent(ctx, 'orphan', { meta: { parentSession: SessionId('missing-parent'), origin: 'subagent' }, }) - expect(selectable(ctx, orphan)).toBe(false) + expect(selectable(ctx, orphan)).toBe(true) const enabledSeed = Session.create(SessionId('enabled-seed')) enabledSeed.append('subagent/model-selection-policy', { allowedModels: ALLOWED_MODELS }) @@ -354,11 +382,11 @@ describe('SubagentModelSelectionConfig', () => { allowedModels: ALLOWED_MODELS, }) const resumedEmpty = await createAgent(ctx, 'resumed-empty', { seed: [] }) - expect(selectable(ctx, resumedEmpty)).toBe(false) + expect(selectable(ctx, resumedEmpty)).toBe(true) expect(subagentModelSelectionPolicy(ctx.sessionProjections, resumedEmpty.session)).toBeUndefined() const resumedDisabled = await createAgent(ctx, 'resumed-disabled', { seed: oldSeed.snapshotEvents() }) - expect(selectable(ctx, resumedDisabled)).toBe(false) + expect(selectable(ctx, resumedDisabled)).toBe(true) expect(subagentModelSelectionPolicy(ctx.sessionProjections, resumedDisabled.session)).toBeUndefined() await ctx.fiber.dispose() }) diff --git a/packages/subagent/tool-subagent/tests/model-selection.spec.ts b/packages/subagent/tool-subagent/tests/model-selection.spec.ts index 5182ecfcc0..4af0606e66 100644 --- a/packages/subagent/tool-subagent/tests/model-selection.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection.spec.ts @@ -1,7 +1,7 @@ import { mountWorkingDirectoryFixture } from '../../subagent/tests/working-directory-fixture.ts' import { describe, expect, it, vi } from 'vitest' import { Context } from '@deepseek-ai/cordis' -import { ReasoningEffortId } from '@deepseek-ai/dsh-llm' +import { ReasoningEffortId, ToolCallId } from '@deepseek-ai/dsh-llm' import SystemPrompt from '@deepseek-ai/dsh-system-prompt' import ToolRuntime from '@deepseek-ai/dsh-tools' import type { Agent } from '@deepseek-ai/dsh-agent' @@ -132,13 +132,22 @@ describe('dsh-tool-subagent model selection', () => { expect(definition?.description).not.toContain('beta') }) - it('hides and rejects route fields when selection is disabled', async () => { - const ctx = await setup({ provider: 'mock' }) + it('rejects route arguments for a backend without Agent options support', async () => { + const ctx = await setup({ provider: 'mock' }, { capabilities: { agentOptions: false } }) + const result = await callSubagent(ctx, { + description: 'unsupported route', prompt: 'do it', provider: 'alpha', model: 'parent-model', + }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('model selection is disabled for this tool instance') + }) + + it('keeps route fields visible but rejects other routes when selection is disabled', async () => { + const ctx = await setup({ provider: 'mock', parentAgentOptions: { provider: 'alpha', model: 'parent-model' } }) const schema = ctx.tools.schemas().find(entry => entry.name === 'subagent')! const props = (schema.parameters as { properties?: Record }).properties ?? {} - expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'prompt']) - expect(schema.description).not.toContain('list_subagent_models') - expect(ctx.tools.get('list_subagent_models')).toBeUndefined() + expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'model', 'prompt', 'provider', 'reasoning_effort']) + expect(schema.description).toContain('list_subagent_models') + expect(ctx.tools.get('list_subagent_models')).toBeDefined() const result = await callSubagent(ctx, { description: 'forced route', @@ -147,7 +156,43 @@ describe('dsh-tool-subagent model selection', () => { model: 'fast-model', }) expect(result.isError).toBe(true) - expect(text(result)).toContain('child model selection is disabled for this tool instance') + expect(text(result)).toContain('is not allowed for this Session') + }) + + it('discovers only the parent route by default and permits its explicit route and effort', async () => { + const requests: SubagentStartRequest[] = [] + const ctx = await setup({ + provider: 'mock', + parentAgentOptions: parentWithRoute().options, + }, { onStart: (request) => { requests.push(request) } }) + ctx.llm.registerAdapter(['alpha', 'other-provider'], new MockAdapter([], REASONING)) + const parent = modelSelectionSetupAgent(ctx) + const discover = (args: unknown) => ctx.tools.execute({ + callId: ToolCallId('default-model-discovery'), name: 'list_subagent_models', + arguments: args, agent: parent, signal: new AbortController().signal, + }) + expect(text(await discover({}))).toContain('alpha') + expect(text(await discover({}))).not.toContain('other-provider') + expect(text(await discover({ provider: 'alpha' }))).toContain('alpha/parent-model') + expect(text(await discover({ provider: 'alpha', model: 'parent-model' }))).toContain('low') + expect((await discover({ provider: 'alpha', model: 'other-model' })).isError).toBe(true) + + const omitted = await callSubagent(ctx, { description: 'inherit', prompt: 'work' }) + expect(omitted.isError, text(omitted)).toBe(false) + expect(requests[0]?.agentOptions).toBeUndefined() + const explicit = await callSubagent(ctx, { + description: 'same route', prompt: 'work', provider: 'alpha', model: 'parent-model', + }) + expect(explicit.isError, text(explicit)).toBe(false) + expect(requests[1]?.agentOptions).toEqual({ provider: 'alpha', model: 'parent-model' }) + const effort = await callSubagent(ctx, { description: 'effort', prompt: 'work', reasoning_effort: 'low' }) + expect(effort.isError, text(effort)).toBe(false) + expect(requests[2]?.agentOptions).toEqual({ reasoningEffort: 'low' }) + const rejected = await callSubagent(ctx, { + description: 'other route', prompt: 'work', provider: 'alpha', model: 'other-model', + }) + expect(rejected.isError).toBe(true) + expect(requests).toHaveLength(3) }) it('rejects enabled model selection when the provider cannot apply Agent options', async () => { diff --git a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts index ea090fa0a3..5993003d03 100644 --- a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts +++ b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts @@ -102,7 +102,10 @@ describe('dsh-tool-subagent', () => { expect(Object.keys(props).sort()).toEqual([ 'cwd', 'description', + 'model', 'prompt', + 'provider', + 'reasoning_effort', ]) expect(schema!.description).not.toContain('job_output') }) @@ -510,7 +513,7 @@ describe('dsh-tool-subagent', () => { const schema = ctx.tools.schemas().find(s => s.name === 'subagent')! expect(schema.description).toContain('inherits this conversation') expect(schema.description).not.toContain('does not see this conversation') - expect(schema.description).not.toContain('can prevent provider-side reuse of the inherited conversation prefix') + expect(schema.description).toContain('can prevent provider-side reuse of the inherited conversation prefix') const props = (schema.parameters as { properties: Record }).properties expect(props['prompt']!.description).toContain('completed turns') }) @@ -760,7 +763,7 @@ describe('dsh-tool-subagent local activation', () => { { description: 'continuable work', prompt: 'dig in' }, { agent: parent }, ) - expect(started.isError).toBe(false) + expect(started.isError, text(started)).toBe(false) const match = /^started subagent (\S+)$/.exec(text(started)) expect(match).not.toBeNull() const [, childId] = match! @@ -828,7 +831,7 @@ describe('dsh-tool-subagent local activation', () => { const [failed, succeeded] = await Promise.all([cancelledResult, survivingResult]) expect(preparationCount).toBe(2) expect(failed.isError).toBe(true) - expect(succeeded.isError).toBe(false) + expect(succeeded.isError, text(succeeded)).toBe(false) expect(cancelledChildId).toBeDefined() expect(survivingChildId).toBeDefined() expect(ctx.agents.get(cancelledChildId!)).toBeUndefined() diff --git a/packages/subagent/tool-subagent/tsdown.config.ts b/packages/subagent/tool-subagent/tsdown.config.ts index 4a467f2c5e..8cdb8ec132 100644 --- a/packages/subagent/tool-subagent/tsdown.config.ts +++ b/packages/subagent/tool-subagent/tsdown.config.ts @@ -15,4 +15,5 @@ const entry = (path: string) => ({ export default defineConfig([ entry('lib/types/index.js'), entry('lib/types/model-selection-settings.js'), + entry('lib/types/list-models.js'), ]) diff --git a/scripts/check-workspace-constraints.ts b/scripts/check-workspace-constraints.ts index c670b46f2a..a15d3c071d 100644 --- a/scripts/check-workspace-constraints.ts +++ b/scripts/check-workspace-constraints.ts @@ -195,9 +195,8 @@ const packageFileExtras: Readonly> = { '@deepseek-ai/dsh-experimental-inspector': ['lib/client.*.js', 'lib/worker.js', 'lib/devtools/**'], // Creator's composition guidance travels with the declaration package. '@deepseek-ai/dsh-agent-preset': ['skills'], - // The Web Host mounts the default-off settings owner independently of each - // Agent-scoped delegation-tool instance. - '@deepseek-ai/dsh-tool-subagent': ['lib/model-selection-settings.js'], + // Settings and shared discovery have independent composition entries. + '@deepseek-ai/dsh-tool-subagent': ['lib/list-models.js', 'lib/model-selection-settings.js'], // The JSONL backend resolves its private verification Worker relative to // import.meta.url; it is shipped without a public package subpath. '@deepseek-ai/dsh-session-persistence-jsonl': ['lib/worker.cjs'], diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index 14d4630c98..fa03619cba 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -73,7 +73,7 @@ import * as PluginManagerTools from '@deepseek-ai/dsh-plugin-manager/tools' import SandboxPolicy from '@deepseek-ai/dsh-sandbox-policy' import McpResources from '@deepseek-ai/dsh-mcp-resources' import * as ToolSubagent from '@deepseek-ai/dsh-tool-subagent' -import { registerListSubagentModels } from '../packages/subagent/tool-subagent/src/list-models.ts' +import * as SubagentModelDiscovery from '../packages/subagent/tool-subagent/src/list-models.ts' import * as ToolWeb from '@deepseek-ai/dsh-tool-web' import WorkflowEngine from '@deepseek-ai/dsh-workflow' import type { WorkflowRun, WorkflowStartRequest } from '@deepseek-ai/dsh-workflow' @@ -519,10 +519,10 @@ const TOOL_PACKAGES: ToolPackage[] = [ await ctx.plugin(LlmRuntime) registerCatalogSubagentProvider(ctx, 'mock') await ctx.plugin(ToolSubagent, { provider: 'mock' }) - registerListSubagentModels(ctx, { routes: [{ provider: 'mock', model: 'mock' }] }) + await ctx.plugin(SubagentModelDiscovery) }, note: - 'The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion available in an enabled Session. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` remains fixed-route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', + 'The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` uses that same policy, and without an enabled policy explicit selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', }, { pkg: '@deepseek-ai/dsh-tool-subagent-control', diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/result.json b/scripts/snapshots/python-sdk-single-exe/advanced/result.json index 8c9eb58af3..bc0bfd16a4 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/result.json +++ b/scripts/snapshots/python-sdk-single-exe/advanced/result.json @@ -207,6 +207,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -303,6 +304,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -1725,6 +1727,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -1851,6 +1854,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -2873,6 +2877,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -3631,6 +3636,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "snapshot_double", "subagent", "workflow", diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl index a169f4f81a..2c5e9c5f05 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl @@ -13,7 +13,7 @@ {"type":"user/message","data":{"id":"{{messageId}}","role":"user","content":[{"type":"text","text":"Serial agent creation completed."}],"source":{"kind":"plugin:serial-created-fixture"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Reply with exactly DIRECT_CHILD_OK and nothing else."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).\n\nYou are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."},{"name":"subagent:delegation","text":"You are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Reply with exactly DIRECT_CHILD_OK and","messageSeqs":[12],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{child-1}}","sessionFormatVersion":4,"throughSeq":16}} diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl index 88aaf9d5e2..5ec93b7783 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl @@ -13,7 +13,7 @@ {"type":"user/message","data":{"id":"{{messageId}}","role":"user","content":[{"type":"text","text":"Serial agent creation completed."}],"source":{"kind":"plugin:serial-created-fixture"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Reply with exactly WORKFLOW_CHILD_OK and nothing else."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).\n\nYou are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."},{"name":"subagent:delegation","text":"You are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Reply with exactly WORKFLOW_CHILD_OK and","messageSeqs":[12],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{child-2}}","sessionFormatVersion":4,"throughSeq":16}} diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl index 3a941d0115..b6dd8f6a4a 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl @@ -13,12 +13,12 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Run the advanced packaged-runtime snapshot scenario."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"id":"{{messageId}}","role":"user","content":[{"type":"image","attachment":{"attachmentId":"sha256:b1ff9c8ea3a780bad09b346c423d2d0e46815926879b18e841d928376a946640","mediaType":"image/png","width":1,"height":1,"bytes":69}}],"source":{"kind":"python-snapshot-image-offload"}},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Run the advanced packaged-runtime snapsh","messageSeqs":[11],"source":{"kind":"fallback"}}} {"type":"assistant/attempt","data":{"turn":1,"step":1,"stream":[{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"error","failure":{"code":"IMAGE_OFFLOAD_REQUIRED","message":"authored image budget failure","offloadImages":1}}}}]}} {"type":"image/offload","data":{"targets":[{"seq":13,"imageIndexes":[0]}]}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"series"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"series"}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{parent}}","sessionFormatVersion":4,"throughSeq":19}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-native","name":"snapshot_double","arguments":"{\"value\": 21}"}],"source":{"kind":"model","provider":"deepseek-official","model":"smoke-model","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"}]}},"id":"{{messageId}}"},"usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"advanced-native","name":"snapshot_double","args":[""]},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"advanced-native","args":["{\"value\": 21}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-native","name":"snapshot_double","arguments":"{\"value\": 21}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"}]}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"advanced-native","name":"snapshot_double","arguments":"{\"value\": 21}"}} diff --git a/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl index c679ff36b7..fc8d44ed52 100644 --- a/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{messageId}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Run three todo updates; the scheduler will reject the second call."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","subagent","todo_write","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","subagent","todo_write","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Run three todo updates; the","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"scheduler-complete","name":"todo_write","arguments":"{\"todos\": []}"},{"type":"tool-call","id":"scheduler-fail","name":"todo_write","arguments":"{\"todos\": []}"},{"type":"tool-call","id":"scheduler-unstarted","name":"todo_write","arguments":"{\"todos\": []}"}],"source":{"kind":"model","provider":"deepseek-official","model":"smoke-model","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"},{"type":"tool-call"},{"type":"tool-call"}]}},"id":"{{messageId}}"},"usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"scheduler-complete","name":"todo_write","args":[""]},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"scheduler-complete","args":["{\"todos\": []}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"scheduler-complete","name":"todo_write","arguments":"{\"todos\": []}"}}},{"type":"chunk","time":0,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":1,"dt":[],"id":"scheduler-fail","name":"todo_write","args":[""]},{"type":"tool-call-chunks","time0":0,"index":1,"dt":[],"id":"scheduler-fail","args":["{\"todos\": []}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"scheduler-fail","name":"todo_write","arguments":"{\"todos\": []}"}}},{"type":"chunk","time":0,"chunk":{"type":"block-start","index":2,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":2,"dt":[],"id":"scheduler-unstarted","name":"todo_write","args":[""]},{"type":"tool-call-chunks","time0":0,"index":2,"dt":[],"id":"scheduler-unstarted","args":["{\"todos\": []}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":2,"block":{"type":"tool-call","id":"scheduler-unstarted","name":"todo_write","arguments":"{\"todos\": []}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"},{"type":"tool-call"},{"type":"tool-call"}]}}}]},"surfaceOp":"append"} diff --git a/snapshots/sdk/bash-tool/tool-schemas.expected.json b/snapshots/sdk/bash-tool/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/bash-tool/tool-schemas.expected.json +++ b/snapshots/sdk/bash-tool/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl b/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl index bfec6f1e36..b497f77dbd 100644 --- a/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl +++ b/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl @@ -6,7 +6,7 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"system/message","seq":7,"time":0,"data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":8,"time":0,"data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":9,"time":0,"data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/context","seq":11,"time":0,"data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"session/title","seq":12,"time":0,"data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":13,"time":0,"data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} @@ -14,14 +14,14 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/result","seq":15,"time":0,"data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_task_read"},"toolCallId":"call_task_read","content":[{"type":"text","text":"{{cwd}}/task.txt\nfile\n\n1: Reply with the single word DONE after reading this file.\n\n(End of file - total 1 lines)\n"}],"isError":false,"id":"{{sessionId}}"},"meta":{"path":"{{cwd}}/task.txt","offset":1,"lines":[{"number":1,"text":"Reply with the single word DONE after reading this file."}],"totalLines":1}},"sourceEventSeqs":[14],"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/end","seq":16,"time":0,"data":{"turn":1,"step":1}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/start","seq":17,"time":0,"data":{"turn":1,"step":2}}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":18,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":18,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":19,"time":0,"data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":18},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":20,"time":0,"data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/call","seq":21,"time":0,"data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/result","seq":22,"time":0,"data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{sessionId}}"}},"sourceEventSeqs":[21],"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/end","seq":23,"time":0,"data":{"turn":1,"step":2}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/start","seq":24,"time":0,"data":{"turn":1,"step":3}}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":25,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":25,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":26,"time":0,"data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{sessionId}}"}},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":27,"time":0,"data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":0,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/end","seq":28,"time":0,"data":{"turn":1,"step":3}}}} diff --git a/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl b/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl index aee79405da..8472750d16 100644 --- a/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl +++ b/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790049391235,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391235,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790049391235,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -17,14 +17,14 @@ {"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_task_read"},"toolCallId":"call_task_read","content":[{"type":"text","text":"{{cwd}}/task.txt\nfile\n\n1: Reply with the single word DONE after reading this file.\n\n(End of file - total 1 lines)\n"}],"isError":false,"id":"{{message:5}}"},"meta":{"path":"{{cwd}}/task.txt","offset":1,"lines":[{"number":1,"text":"Reply with the single word DONE after reading this file."}],"totalLines":1}},"sourceEventSeqs":[14],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:6}}"},"headerSeq":18},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:7}}"},"stream":[{"type":"chunk","time":1790049391255,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391255,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790049391255,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790049391255,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:8}}"}},"sourceEventSeqs":[21],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:9}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:10}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790049391270,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790049391270,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790049391270,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/sdk/persistent-tools/cordis.yml b/snapshots/sdk/persistent-tools/cordis.yml index 6262d2fb44..e2ee87e8a7 100644 --- a/snapshots/sdk/persistent-tools/cordis.yml +++ b/snapshots/sdk/persistent-tools/cordis.yml @@ -84,6 +84,10 @@ name: '@deepseek-ai/dsh-tool-subagent-control' disabled: true +- id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' + disabled: true + - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' disabled: true diff --git a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json index ed6f50a025..17dff0816f 100644 --- a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_dsh_sdk", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider's route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or this provider's route defaults." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or this provider's route defaults." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to use a compatible configured effort or the selected model's default." } }, "required": [ @@ -389,7 +430,7 @@ }, { "name": "subagent_dsh_sdk_second", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider's route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -404,6 +445,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or this provider's route defaults." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or this provider's route defaults." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to use a compatible configured effort or the selected model's default." } }, "required": [ @@ -414,7 +467,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -429,6 +482,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json index ce1cf8dfd5..87515e6d78 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json @@ -262,7 +262,7 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its advertised models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", "parameters": { "type": "object", "properties": { @@ -272,7 +272,7 @@ }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's advertised models." + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." } } } @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -408,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/text-turn/tool-schemas.expected.json b/snapshots/sdk/text-turn/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/sdk/text-turn/tool-schemas.expected.json +++ b/snapshots/sdk/text-turn/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/tool-error-details/notifications.expected.jsonl b/snapshots/sdk/tool-error-details/notifications.expected.jsonl index e6d6e7fde2..8d30d3d16c 100644 --- a/snapshots/sdk/tool-error-details/notifications.expected.jsonl +++ b/snapshots/sdk/tool-error-details/notifications.expected.jsonl @@ -6,7 +6,7 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"system/message","seq":7,"time":0,"data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":8,"time":0,"data":{"content":[{"type":"text","text":"Exercise a native denial and a caught PTC inner denial, then reply ERROR_DETAILS_OK."}],"source":{"kind":"user"},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":9,"time":0,"data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/context","seq":11,"time":0,"data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"session/title","seq":12,"time":0,"data":{"title":"Exercise a native denial and","messageSeqs":[8],"source":{"kind":"fallback"}}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":13,"time":0,"data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"denied-native","name":"bash","args":["{\"command\":\"echo AUTO_DENIED\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} @@ -16,7 +16,7 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/start","seq":17,"time":0,"data":{"turn":1,"step":2}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"system/message","seq":18,"time":0,"data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":19,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["run_code"]},"reason":"change","startsSeries":true}}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":20,"time":0,"data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":19},"surfaceOp":"append"}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":20,"time":0,"data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"list_subagent_models"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":19},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":21,"time":0,"data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"denied-code","name":"run_code","args":["{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/call","seq":22,"time":0,"data":{"turn":1,"step":2,"callId":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/ptc-dispatch-start","seq":23,"time":0,"data":{"rootCallId":"denied-code","parentCallId":"denied-code","subCallId":"denied-code:ptc:1","name":"bash","arguments":{"command":"echo AUTO_DENIED"}}}}} diff --git a/snapshots/sdk/tool-error-details/session.v4.jsonl b/snapshots/sdk/tool-error-details/session.v4.jsonl index 871095019b..806aa48ff4 100644 --- a/snapshots/sdk/tool-error-details/session.v4.jsonl +++ b/snapshots/sdk/tool-error-details/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Exercise a native denial and a caught PTC inner denial, then reply ERROR_DETAILS_OK."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Exercise a native denial and","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"stream":[{"type":"chunk","time":1791315723125,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315723125,"index":0,"dt":[],"id":"denied-native","name":"bash","args":["{\"command\":\"echo AUTO_DENIED\"}"]},{"type":"chunk","time":1791315723125,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}}},{"type":"chunk","time":1791315723125,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -19,7 +19,7 @@ {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["run_code"]},"reason":"change","startsSeries":true}} -{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} +{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"list_subagent_models"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1791315723151,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315723151,"index":0,"dt":[],"id":"denied-code","name":"run_code","args":["{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"]},{"type":"chunk","time":1791315723151,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}}},{"type":"chunk","time":1791315723151,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}} {"type":"tool/ptc-dispatch-start","data":{"rootCallId":"denied-code","parentCallId":"denied-code","subCallId":"denied-code:ptc:1","name":"bash","arguments":{"command":"echo AUTO_DENIED"}}} diff --git a/snapshots/sdk/tool-error-details/system-prompt.expected.md b/snapshots/sdk/tool-error-details/system-prompt.expected.md index 9070405e5b..b0a8db4970 100644 --- a/snapshots/sdk/tool-error-details/system-prompt.expected.md +++ b/snapshots/sdk/tool-error-details/system-prompt.expected.md @@ -173,6 +173,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -199,7 +206,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -207,8 +214,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -216,6 +229,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -440,6 +459,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/sdk/tool-error-details/tool-schemas.expected.json b/snapshots/sdk/tool-error-details/tool-schemas.expected.json index e150d82593..5e9eaf2f00 100644 --- a/snapshots/sdk/tool-error-details/tool-schemas.expected.json +++ b/snapshots/sdk/tool-error-details/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/advanced-toolchain/session.v4.jsonl b/snapshots/session/advanced-toolchain/session.v4.jsonl index c0f1a22b8d..b9591d4802 100644 --- a/snapshots/session/advanced-toolchain/session.v4.jsonl +++ b/snapshots/session/advanced-toolchain/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Run this advanced flow exactly once: list runtime inspect providers; discover providers through run_code when available; delegate once to a direct spawn child; run one workflow that delegates to another spawn child; list providers again; then reply with exactly ADVANCED_HEADLESS_OK."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Run this advanced flow exactly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-inspect-before","name":"cordis_inspect_list","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1791315715021,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315715022,"index":0,"dt":[],"id":"advanced-inspect-before","name":"cordis_inspect_list","args":["{}"]},{"type":"chunk","time":1791315715022,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-inspect-before","name":"cordis_inspect_list","arguments":"{}"}}},{"type":"chunk","time":1791315715022,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1791315715022,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -19,7 +19,7 @@ {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["run_code"]},"reason":"change","startsSeries":true}} -{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"cordis_inspect_list"},{"type":"tool-removal","toolName":"cordis_inspect_query"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} +{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"cordis_inspect_list"},{"type":"tool-removal","toolName":"cordis_inspect_query"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"list_subagent_models"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-code","name":"run_code","arguments":"{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1791315715055,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315715055,"index":0,"dt":[],"id":"advanced-code","name":"run_code","args":["{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"]},{"type":"chunk","time":1791315715055,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-code","name":"run_code","arguments":"{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"}}},{"type":"chunk","time":1791315715055,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1791315715055,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"advanced-code","name":"run_code","arguments":"{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"}} {"type":"tool/ptc-dispatch-start","data":{"rootCallId":"advanced-code","parentCallId":"advanced-code","subCallId":"advanced-code:ptc:1","name":"cordis_inspect_list","arguments":{}}} @@ -28,8 +28,8 @@ {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} {"type":"system/message","data":{"turn":1,"step":3,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:10}}"}},"sourceEventSeqs":[18],"surfaceOp":{"op":"replace","startSeq":18,"endSeq":18}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} -{"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"cordis_inspect_list"},{"type":"tool-addition","toolName":"cordis_inspect_query"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"},{"type":"tool-removal","toolName":"run_code"}],"role":"developer","id":"{{message:11}}"},"headerSeq":29},"surfaceOp":"append"} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} +{"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"cordis_inspect_list"},{"type":"tool-addition","toolName":"cordis_inspect_query"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"list_subagent_models"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"},{"type":"tool-removal","toolName":"run_code"}],"role":"developer","id":"{{message:11}}"},"headerSeq":29},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-direct-child","name":"subagent","arguments":"{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:12}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1791315715200,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315715200,"index":0,"dt":[],"id":"advanced-direct-child","name":"subagent","args":["{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"]},{"type":"chunk","time":1791315715200,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-direct-child","name":"subagent","arguments":"{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"}}},{"type":"chunk","time":1791315715200,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1791315715200,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":3,"callId":"advanced-direct-child","name":"subagent","arguments":"{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"}} {"type":"subagent/catalog","data":{"version":0,"childId":"{{session:2}}","childCreatedAt":1791315715221,"mode":"continuable","label":"Check direct child"}} diff --git a/snapshots/session/advanced-toolchain/system-prompt.expected.md b/snapshots/session/advanced-toolchain/system-prompt.expected.md index cf4dcd9c3f..68033a12e2 100644 --- a/snapshots/session/advanced-toolchain/system-prompt.expected.md +++ b/snapshots/session/advanced-toolchain/system-prompt.expected.md @@ -186,6 +186,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -212,7 +219,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -220,8 +227,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -229,6 +242,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -455,6 +474,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/advanced-toolchain/tool-schemas.expected.json b/snapshots/session/advanced-toolchain/tool-schemas.expected.json index cb349d3ca4..c0890029de 100644 --- a/snapshots/session/advanced-toolchain/tool-schemas.expected.json +++ b/snapshots/session/advanced-toolchain/tool-schemas.expected.json @@ -301,6 +301,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -380,7 +397,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -395,6 +412,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -405,7 +434,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -420,6 +449,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1014,6 +1055,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1093,7 +1151,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -1108,6 +1166,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1118,7 +1188,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -1133,6 +1203,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/agent-instructions/tool-schemas.expected.json b/snapshots/session/agent-instructions/tool-schemas.expected.json index 00e71002a7..472458e753 100644 --- a/snapshots/session/agent-instructions/tool-schemas.expected.json +++ b/snapshots/session/agent-instructions/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -893,6 +934,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -972,7 +1030,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -987,6 +1045,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -997,7 +1067,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -1012,6 +1082,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/background-confinement-failure/tool-schemas.expected.json b/snapshots/session/background-confinement-failure/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/background-confinement-failure/tool-schemas.expected.json +++ b/snapshots/session/background-confinement-failure/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/both-mode-turn/system-prompt.expected.md b/snapshots/session/both-mode-turn/system-prompt.expected.md index 329e451ce8..703d69b43f 100644 --- a/snapshots/session/both-mode-turn/system-prompt.expected.md +++ b/snapshots/session/both-mode-turn/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -407,6 +426,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json index cd473bea70..a8ef352d5e 100644 --- a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json @@ -300,6 +300,23 @@ ] } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "mcp__chrome-devtools-mcp__take_screenshot", "description": "Take a screenshot of the page or element.", @@ -446,7 +463,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -461,6 +478,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -471,7 +500,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -486,6 +515,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json index 2d24d7c17d..3d46faeef4 100644 --- a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json @@ -300,6 +300,23 @@ ] } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "mcp__playwright-mcp__browser_take_screenshot", "description": "Take a screenshot of the current page. You can't perform actions based on the screenshot, use browser_snapshot for actions.", @@ -448,7 +465,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -463,6 +480,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -473,7 +502,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -488,6 +517,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json index e1acbc8206..33c0bdbb5a 100644 --- a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json +++ b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -552,7 +569,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -567,6 +584,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -577,7 +606,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -592,6 +621,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/claude-code-mods/tool-schemas.expected.json b/snapshots/session/claude-code-mods/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/claude-code-mods/tool-schemas.expected.json +++ b/snapshots/session/claude-code-mods/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/compaction-output-reserve/cordis.snapshot.yml b/snapshots/session/compaction-output-reserve/cordis.snapshot.yml index 3dce48221f..15f69e1913 100644 --- a/snapshots/session/compaction-output-reserve/cordis.snapshot.yml +++ b/snapshots/session/compaction-output-reserve/cordis.snapshot.yml @@ -1,5 +1,5 @@ # With 1,800 tokens of headroom, a 1,000-token output reservation lowers the -# trigger from 6,782 to 5,782, between the history and its condensed replacement. +# trigger from 7,470 to 6,470, between the history and its condensed replacement. - id: agent-instructions disabled: true @@ -31,7 +31,7 @@ name: DeepSeek models: - id: deepseek-v4-flash - contextWindow: 8582 + contextWindow: 9270 defaultMaxTokens: 1000 - id: compaction-working-directory name: '../../../packages/test-support/session-snapshot/tests/fixtures/compaction-working-directory.ts' diff --git a/snapshots/session/compaction-output-reserve/cordis.yml b/snapshots/session/compaction-output-reserve/cordis.yml index a8d2088083..ea561fc74f 100644 --- a/snapshots/session/compaction-output-reserve/cordis.yml +++ b/snapshots/session/compaction-output-reserve/cordis.yml @@ -7,7 +7,7 @@ config: models: - id: deepseek-v4-flash - contextWindow: 8582 + contextWindow: 9270 maxTokens: 1000 - id: compaction-basic diff --git a/snapshots/session/compaction-output-reserve/session.v4.jsonl b/snapshots/session/compaction-output-reserve/session.v4.jsonl index 4b9830d6ce..ed3f0c8a16 100644 --- a/snapshots/session/compaction-output-reserve/session.v4.jsonl +++ b/snapshots/session/compaction-output-reserve/session.v4.jsonl @@ -11,7 +11,7 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Establish a durable compaction premise before continuing. Record every part of this historical evidence: the snapshot uses keyless replay; persistence uses JSONL; the assembled headless application loads its real Cordis composition; model-visible inputs remain logged; tool calls and results remain paired and ordered; output reservation reduces the message budget before proactive compaction; compaction opens with compaction/start and closes with compaction/end; a successful auxiliary summary records compaction/summary source metadata; the replacement surface shadows only an older balanced range; the checkpoint remains smaller than the history it replaces; the newest tool result remains verbatim; the next request sees that checkpoint; the final response proves the same turn continued; deterministic snapshot evidence stays separate from the live-provider smoke; and no external API key is needed. Emit one alpha marker through bash, then finish the task after proactive compaction with the exact words PROACTIVE COMPACTION COMPLETE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"/tmp\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"/tmp\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":1000},"adapterDefaults":{"maxTokens":true},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":8582}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9270}} {"type":"session/title","data":{"title":"Establish a durable compaction premise","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":24,"outputTokens":6},"stream":[{"type":"chunk","time":1791185986560,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791185986560,"index":0,"dt":[],"id":"call_compaction_marker","name":"bash","args":["{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"]},{"type":"chunk","time":1791185986560,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}}},{"type":"chunk","time":1791185986560,"chunk":{"type":"usage","usage":{"inputTokens":24,"outputTokens":6}}},{"type":"chunk","time":1791185986560,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}} diff --git a/snapshots/session/compaction-recovery/tool-schemas.expected.json b/snapshots/session/compaction-recovery/tool-schemas.expected.json index 00e71002a7..472458e753 100644 --- a/snapshots/session/compaction-recovery/tool-schemas.expected.json +++ b/snapshots/session/compaction-recovery/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -893,6 +934,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -972,7 +1030,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -987,6 +1045,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -997,7 +1067,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -1012,6 +1082,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml b/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml index 765a3f7e8a..76355f3bb4 100644 --- a/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml +++ b/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml @@ -1,5 +1,5 @@ # The summary output cap inherits the 1,800-token headroom. A 1,000-token -# request output reservation lowers the proactive trigger to 5,782 tokens. +# request output reservation lowers the proactive trigger to 6,470 tokens. - id: agent-instructions disabled: true @@ -30,7 +30,7 @@ name: DeepSeek models: - id: deepseek-v4-flash - contextWindow: 8582 + contextWindow: 9270 defaultMaxTokens: 1000 - id: compaction-working-directory name: '../../../packages/test-support/session-snapshot/tests/fixtures/compaction-working-directory.ts' diff --git a/snapshots/session/compaction-summary-headroom/cordis.yml b/snapshots/session/compaction-summary-headroom/cordis.yml index f65c90899e..da55c9afc7 100644 --- a/snapshots/session/compaction-summary-headroom/cordis.yml +++ b/snapshots/session/compaction-summary-headroom/cordis.yml @@ -7,7 +7,7 @@ config: models: - id: deepseek-v4-flash - contextWindow: 8582 + contextWindow: 9270 maxTokens: 1000 - id: compaction-basic diff --git a/snapshots/session/compaction-summary-headroom/session.v4.jsonl b/snapshots/session/compaction-summary-headroom/session.v4.jsonl index 2891a589e1..7d2da30924 100644 --- a/snapshots/session/compaction-summary-headroom/session.v4.jsonl +++ b/snapshots/session/compaction-summary-headroom/session.v4.jsonl @@ -11,7 +11,7 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Establish a durable compaction premise before continuing. Record every part of this historical evidence: the snapshot uses keyless replay; persistence uses JSONL; the assembled headless application loads its real Cordis composition; model-visible inputs remain logged; tool calls and results remain paired and ordered; output reservation reduces the message budget before proactive compaction; compaction opens with compaction/start and closes with compaction/end; a successful auxiliary summary records compaction/summary source metadata; the replacement surface shadows only an older balanced range; the checkpoint remains smaller than the history it replaces; the newest tool result remains verbatim; the next request sees that checkpoint; the final response proves the same turn continued; deterministic snapshot evidence stays separate from the live-provider smoke; and no external API key is needed. Emit one alpha marker through bash, then finish the task after proactive compaction with the exact words PROACTIVE COMPACTION COMPLETE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"/tmp\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"/tmp\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":1000},"adapterDefaults":{"maxTokens":true},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":8582}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9270}} {"type":"session/title","data":{"title":"Establish a durable compaction premise","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":24,"outputTokens":6},"stream":[{"type":"chunk","time":1791187853875,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791187853875,"index":0,"dt":[],"id":"call_compaction_marker","name":"bash","args":["{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"]},{"type":"chunk","time":1791187853875,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}}},{"type":"chunk","time":1791187853875,"chunk":{"type":"usage","usage":{"inputTokens":24,"outputTokens":6}}},{"type":"chunk","time":1791187853875,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}} diff --git a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json index 4b839adb31..9a7dc19de7 100644 --- a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json @@ -300,6 +300,23 @@ ] } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "mcp__cua-driver-mcp__disconnect", "description": "Disconnect the fixture.", @@ -425,7 +442,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -440,6 +457,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -450,7 +479,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -465,6 +494,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json index 802ed2fe0d..a95daa24a6 100644 --- a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json @@ -306,6 +306,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -385,7 +402,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -400,6 +417,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -410,7 +439,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -425,6 +454,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json index 5d880fdde7..da58565d4e 100644 --- a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json +++ b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json @@ -301,6 +301,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -380,7 +397,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -395,6 +412,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -405,7 +434,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -420,6 +449,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl b/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl index cff0e5a329..53e2ccd74b 100644 --- a/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl +++ b/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","systemPromptUpdate":"in-history"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790149644504,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644504,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790149644504,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -18,14 +18,14 @@ {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1790149644534,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644534,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790149644534,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790149644535,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:9}}"}},"sourceEventSeqs":[22],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:10}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:11}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790149644557,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790149644557,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790149644557,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/session/dynamic-tool-updates/session.v4.jsonl b/snapshots/session/dynamic-tool-updates/session.v4.jsonl index aee79405da..8472750d16 100644 --- a/snapshots/session/dynamic-tool-updates/session.v4.jsonl +++ b/snapshots/session/dynamic-tool-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790049391235,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391235,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790049391235,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -17,14 +17,14 @@ {"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_task_read"},"toolCallId":"call_task_read","content":[{"type":"text","text":"{{cwd}}/task.txt\nfile\n\n1: Reply with the single word DONE after reading this file.\n\n(End of file - total 1 lines)\n"}],"isError":false,"id":"{{message:5}}"},"meta":{"path":"{{cwd}}/task.txt","offset":1,"lines":[{"number":1,"text":"Reply with the single word DONE after reading this file."}],"totalLines":1}},"sourceEventSeqs":[14],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:6}}"},"headerSeq":18},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:7}}"},"stream":[{"type":"chunk","time":1790049391255,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391255,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790049391255,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790049391255,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:8}}"}},"sourceEventSeqs":[21],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:9}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:10}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790049391270,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790049391270,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790049391270,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json index 87bccafb3f..5268b0e9d8 100644 --- a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json +++ b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -893,6 +934,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -980,7 +1038,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -995,6 +1053,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1005,7 +1075,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -1020,6 +1090,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1533,6 +1615,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1612,7 +1711,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -1627,6 +1726,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1637,7 +1748,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -1652,6 +1763,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json +++ b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json index 3fb3980c3b..840733798a 100644 --- a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json +++ b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json @@ -180,6 +180,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -243,7 +260,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -258,6 +275,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -268,7 +297,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -283,6 +312,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/lsp-definition/tool-schemas.expected.json b/snapshots/session/lsp-definition/tool-schemas.expected.json index 76087bb96f..473b6cc111 100644 --- a/snapshots/session/lsp-definition/tool-schemas.expected.json +++ b/snapshots/session/lsp-definition/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "lsp", "description": "Query a language server for precise code navigation. operation is one of goToDefinition, findReferences, goToImplementation, hover. line and character are one-based UTF-16 cursor coordinates. findReferences includes the declaration.", @@ -376,7 +393,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -391,6 +408,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -401,7 +430,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -416,6 +445,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md index a6b0b07911..2776836c5d 100644 --- a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md +++ b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md @@ -162,6 +162,13 @@ interface ToolArgsMap { /** Continuation cursor returned by this server. */ cursor?: string; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -195,7 +202,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -203,8 +210,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -212,6 +225,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -438,6 +457,7 @@ interface ToolOutputMap { })[]; list_mcp_resource_templates: JsonValue; list_mcp_resources: JsonValue; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/mcp-resources/tool-schemas.expected.json b/snapshots/session/mcp-resources/tool-schemas.expected.json index 4e5e9d557c..5a7f33f534 100644 --- a/snapshots/session/mcp-resources/tool-schemas.expected.json +++ b/snapshots/session/mcp-resources/tool-schemas.expected.json @@ -300,6 +300,23 @@ ] } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -400,7 +417,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -415,6 +432,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -425,7 +454,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -440,6 +469,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json index 637c069b74..dec41c176b 100644 --- a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json +++ b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "mcp__fixture__inspect", "description": "Read a fixed window snapshot.", @@ -358,7 +375,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -373,6 +390,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -383,7 +412,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -398,6 +427,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/office-skills/tool-schemas.expected.json b/snapshots/session/office-skills/tool-schemas.expected.json index 79d6891abc..7269632ffd 100644 --- a/snapshots/session/office-skills/tool-schemas.expected.json +++ b/snapshots/session/office-skills/tool-schemas.expected.json @@ -215,6 +215,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -294,7 +311,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -309,6 +326,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -319,7 +348,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -334,6 +363,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/plugin-manager-mcp/session.v4.jsonl b/snapshots/session/plugin-manager-mcp/session.v4.jsonl index 794d19b23f..0036cc1272 100644 --- a/snapshots/session/plugin-manager-mcp/session.v4.jsonl +++ b/snapshots/session/plugin-manager-mcp/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Use plugin_manager with action set_plugin, target include:demo, enabled true to activate the configured MCP entry. Then call its newly discovered mcp__demo__ping tool and report the result. Do not use other tools."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","plugin_manager","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","plugin_manager","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":1000000}} {"type":"session/title","data":{"title":"Use plugin_manager with action set_plugi","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"reasoning","text":"The user wants me to activate a plugin via plugin_manager, then call an MCP tool. Let me do the plugin_manager call first.\n\nNote: they say \"Do not use other tools\" other than plugin_manager and the mcp tool. But I need to first enable it. Let's call plugin_manager set_plugin.\n\nLet me also list plugins first? The instruction says use plugin_manager with action set_plugin, target include:demo, enabled true. Let's just do that."},{"type":"tool-call","id":"call_00_xLhr9wia3A5m7wuSRWwD0436","name":"plugin_manager","arguments":"{\"action\": \"set_plugin\", \"target\": \"include:demo\", \"enabled\": true}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"75d99e21-1d47-4a33-bb88-340a1e84b749"},{"type":"tool-call"}]}},"id":"{{message:4}}"},"usage":{"inputTokens":7220,"outputTokens":174,"cacheReadTokens":0,"cacheWriteTokens":0,"totalTokens":7394},"stream":[{"type":"chunk","time":1789984418848,"chunk":{"type":"block-start","index":0,"blockType":"reasoning"}},{"type":"reasoning-chunks","time0":1789984418848,"index":0,"dt":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"texts":["The"," user"," wants"," me"," to"," activate"," a"," plugin"," via"," plugin","_manager",","," then"," call"," an"," M","CP"," tool","."," Let"," me"," do"," the"," plugin","_manager"," call"," first",".\n\n","Note",":"," they"," say"," \"","Do"," not"," use"," other"," tools","\""," other"," than"," plugin","_manager"," and"," the"," m","cp"," tool","."," But"," I"," need"," to"," first"," enable"," it","."," Let","'s"," call"," plugin","_manager"," set","_","plugin",".\n\n","Let"," me"," also"," list"," plugins"," first","?"," The"," instruction"," says"," use"," plugin","_manager"," with"," action"," set","_","plugin",","," target"," include",":","demo",","," enabled"," true","."," Let","'s"," just"," do"," that","."]},{"type":"chunk","time":1789984418849,"chunk":{"type":"block-end","index":0,"block":{"type":"reasoning","text":"The user wants me to activate a plugin via plugin_manager, then call an MCP tool. Let me do the plugin_manager call first.\n\nNote: they say \"Do not use other tools\" other than plugin_manager and the mcp tool. But I need to first enable it. Let's call plugin_manager set_plugin.\n\nLet me also list plugins first? The instruction says use plugin_manager with action set_plugin, target include:demo, enabled true. Let's just do that."}}},{"type":"chunk","time":1789984418849,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1789984418849,"index":1,"dt":[],"id":"call_00_xLhr9wia3A5m7wuSRWwD0436","name":"plugin_manager","args":[""]},{"type":"tool-call-chunks","time0":1789984418849,"index":1,"dt":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"id":"call_00_xLhr9wia3A5m7wuSRWwD0436","args":["{","\"","action","\"",": ","\"","set","_","plugin","\"",", ","\"","target","\"",": ","\"","include",":","demo","\"",", ","\"","en","abled","\"",": ","true","}"]},{"type":"chunk","time":1789984418849,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"call_00_xLhr9wia3A5m7wuSRWwD0436","name":"plugin_manager","arguments":"{\"action\": \"set_plugin\", \"target\": \"include:demo\", \"enabled\": true}"}}},{"type":"chunk","time":1789984418849,"chunk":{"type":"usage","usage":{"inputTokens":7220,"outputTokens":174,"cacheReadTokens":0,"cacheWriteTokens":0,"totalTokens":7394}}},{"type":"chunk","time":1789984418849,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"75d99e21-1d47-4a33-bb88-340a1e84b749"},{"type":"tool-call"}]}}}]},"surfaceOp":"append"} @@ -18,7 +18,7 @@ {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_mcp_resource_templates","list_mcp_resources","mcp__demo__ping","plugin_manager","read","read_image","read_mcp_resource","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_mcp_resource_templates","list_mcp_resources","list_subagent_models","mcp__demo__ping","plugin_manager","read","read_image","read_mcp_resource","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"list_mcp_resource_templates"},{"type":"tool-addition","toolName":"list_mcp_resources"},{"type":"tool-addition","toolName":"mcp__demo__ping"},{"type":"tool-addition","toolName":"read_mcp_resource"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"reasoning","text":"The plugin was enabled. Now call mcp__demo__ping.\n\nWait — the tool was already listed in my tools before enabling. Let's call it anyway."},{"type":"tool-call","id":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"b1ec1efe-4be6-4836-a10c-e5d6bfdce520"},{"type":"tool-call"}]}},"id":"{{message:8}}"},"usage":{"inputTokens":7035,"outputTokens":59,"cacheReadTokens":768,"cacheWriteTokens":0,"totalTokens":7862},"stream":[{"type":"chunk","time":1789984418998,"chunk":{"type":"block-start","index":0,"blockType":"reasoning"}},{"type":"reasoning-chunks","time0":1789984418998,"index":0,"dt":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"texts":["The"," plugin"," was"," enabled","."," Now"," call"," m","cp","__","demo","__","ping",".\n\n","Wait"," —"," the"," tool"," was"," already"," listed"," in"," my"," tools"," before"," enabling","."," Let","'s"," call"," it"," anyway","."]},{"type":"chunk","time":1789984418998,"chunk":{"type":"block-end","index":0,"block":{"type":"reasoning","text":"The plugin was enabled. Now call mcp__demo__ping.\n\nWait — the tool was already listed in my tools before enabling. Let's call it anyway."}}},{"type":"chunk","time":1789984418998,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1789984418998,"index":1,"dt":[],"id":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","args":[""]},{"type":"tool-call-chunks","time0":1789984418998,"index":1,"dt":[],"id":"call_00_miuKfLCTz5Cb38gdp8kK5186","args":["{}"]},{"type":"chunk","time":1789984418998,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","arguments":"{}"}}},{"type":"chunk","time":1789984418998,"chunk":{"type":"usage","usage":{"inputTokens":7035,"outputTokens":59,"cacheReadTokens":768,"cacheWriteTokens":0,"totalTokens":7862}}},{"type":"chunk","time":1789984418998,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"b1ec1efe-4be6-4836-a10c-e5d6bfdce520"},{"type":"tool-call"}]}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","arguments":"{}"}} diff --git a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json index f44f52640a..bc73a4b1fb 100644 --- a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "plugin_manager", "description": "List plugins or bundles in the current profile, enable or disable them, install a bundle, or remove an installed bundle. Every action requires danger-full-access permission or approval for this call. Approval does not change the session permission mode. Changes affect every session in this profile. List first to obtain exact identifiers. Package installation can execute allowed build scripts. Live profiles apply changes immediately; startup profiles require restart. Incompatible DSH peer dependencies block installation and activation. Version exemptions risk crashes and data loss: warn the user and obtain explicit permission for the exact plugin and runtime versions before granting one.", @@ -400,7 +417,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -415,6 +432,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -425,7 +454,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -440,6 +469,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -994,6 +1035,23 @@ ] } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "mcp__demo__ping", "description": "Replies pong.", @@ -1164,7 +1222,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -1179,6 +1237,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1189,7 +1259,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -1204,6 +1274,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/plugin-manager/tool-schemas.expected.json b/snapshots/session/plugin-manager/tool-schemas.expected.json index 4596dae7a1..47b4c83859 100644 --- a/snapshots/session/plugin-manager/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "plugin_manager", "description": "List plugins or bundles in the current profile, enable or disable them, install a bundle, or remove an installed bundle. Every action requires danger-full-access permission or approval for this call. Approval does not change the session permission mode. Changes affect every session in this profile. List first to obtain exact identifiers. Package installation can execute allowed build scripts. Live profiles apply changes immediately; startup profiles require restart. Incompatible DSH peer dependencies block installation and activation. Version exemptions risk crashes and data loss: warn the user and obtain explicit permission for the exact plugin and runtime versions before granting one.", @@ -400,7 +417,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -415,6 +432,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -425,7 +454,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -440,6 +469,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/product-subagent-both/tool-schemas.expected.json b/snapshots/session/product-subagent-both/tool-schemas.expected.json index 2b36308f92..779cf01085 100644 --- a/snapshots/session/product-subagent-both/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-both/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -464,7 +493,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -479,6 +508,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/product-subagent-codex/tool-schemas.expected.json b/snapshots/session/product-subagent-codex/tool-schemas.expected.json index d9f3cfa9d0..40da027b47 100644 --- a/snapshots/session/product-subagent-codex/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-codex/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -414,7 +443,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -429,6 +458,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json index 11f07197a2..d202776234 100644 --- a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -389,7 +418,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -404,6 +433,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/provider-cwd/tool-schemas.expected.json b/snapshots/session/provider-cwd/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/provider-cwd/tool-schemas.expected.json +++ b/snapshots/session/provider-cwd/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ptc-node-read-only/system-prompt.expected.md b/snapshots/session/ptc-node-read-only/system-prompt.expected.md index 39529ce115..ec62c1bdaa 100644 --- a/snapshots/session/ptc-node-read-only/system-prompt.expected.md +++ b/snapshots/session/ptc-node-read-only/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { @@ -411,6 +430,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-node-workspace/system-prompt.expected.md b/snapshots/session/ptc-node-workspace/system-prompt.expected.md index 39529ce115..ec62c1bdaa 100644 --- a/snapshots/session/ptc-node-workspace/system-prompt.expected.md +++ b/snapshots/session/ptc-node-workspace/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { @@ -411,6 +430,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-python-turn/system-prompt.expected.md b/snapshots/session/ptc-python-turn/system-prompt.expected.md index 23dc2e1494..6f94c095fb 100644 --- a/snapshots/session/ptc-python-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-python-turn/system-prompt.expected.md @@ -287,6 +287,13 @@ class ListAgentsOutput2(TypedDict): parent: NotRequired[str] depth: NotRequired[float] +class ListSubagentModelsArgs(TypedDict): + # Registered LLM provider id. Omit to list providers. + provider: NotRequired[str] + # Exact model id to inspect. Requires provider; omit to list that provider's allowed models. + model: NotRequired[str] + # Additional keys beyond those declared are allowed. + class ReadArgs(TypedDict): # Path to read, resolved by the filesystem backend. file_path: str @@ -370,6 +377,12 @@ class SubagentArgs(TypedDict): description: str # The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. prompt: str + # LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. + provider: NotRequired[str] + # Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. + model: NotRequired[str] + # Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. + reasoning_effort: NotRequired[str] # Additional keys beyond those declared are allowed. class SubagentOutput(TypedDict): @@ -383,6 +396,12 @@ class SubagentForkArgs(TypedDict): description: str # The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. prompt: str + # LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. + provider: NotRequired[str] + # Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. + model: NotRequired[str] + # Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. + reasoning_effort: NotRequired[str] # Additional keys beyond those declared are allowed. class SubagentForkOutput(TypedDict): @@ -535,6 +554,8 @@ class Tools(Protocol): """Read a background job: output since the previous read for stream jobs, or the result of a finished final-output job.""" async def list_agents(self, args: ListAgentsArgs) -> list[ListAgentsOutput1 | ListAgentsOutput2]: """List subagents you started, with their ids, labels, and status. running means it is working; inactive means it is not currently working. You will be notified when a subagent finishes; there is no need to keep checking its status. Use send_message to continue the conversation.""" + async def list_subagent_models(self, args: ListSubagentModelsArgs) -> str: + """Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.""" async def read(self, args: ReadArgs) -> ReadOutput: """Read a UTF-8 text file and return line-numbered content.""" async def read_image(self, args: ReadImageArgs) -> ReadImageOutput: @@ -544,9 +565,9 @@ class Tools(Protocol): async def skill(self, args: SkillArgs) -> SkillOutput: """Load the full instructions for a skill. Call it before acting on a task that names or clearly matches a skill in the session skill catalog.""" async def subagent(self, args: SubagentArgs) -> SubagentOutput: - """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.""" + """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.""" async def subagent_fork(self, args: SubagentForkArgs) -> SubagentForkOutput: - """Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.""" + """Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.""" async def todo_write(self, args: TodoWriteArgs) -> TodoWriteOutput: """Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done.""" async def update_goal(self, args: UpdateGoalArgs) -> UpdateGoalOutput1 | UpdateGoalOutput2: diff --git a/snapshots/session/ptc-read-image/system-prompt.expected.md b/snapshots/session/ptc-read-image/system-prompt.expected.md index f0a3c2426f..62aab7eb38 100644 --- a/snapshots/session/ptc-read-image/system-prompt.expected.md +++ b/snapshots/session/ptc-read-image/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -407,6 +426,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-turn/system-prompt.expected.md b/snapshots/session/ptc-turn/system-prompt.expected.md index 39529ce115..ec62c1bdaa 100644 --- a/snapshots/session/ptc-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-turn/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { @@ -411,6 +430,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json index 95ffc0255b..d33cb92a78 100644 --- a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json +++ b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.1.expected.json b/snapshots/session/ralph-loop/tool-schemas.1.expected.json index 02ebdc4baf..93d4b861a1 100644 --- a/snapshots/session/ralph-loop/tool-schemas.1.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "ralph", "description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.", @@ -402,7 +419,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -417,6 +434,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -427,7 +456,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -442,6 +471,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.2.expected.json b/snapshots/session/ralph-loop/tool-schemas.2.expected.json index 02ebdc4baf..93d4b861a1 100644 --- a/snapshots/session/ralph-loop/tool-schemas.2.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.2.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "ralph", "description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.", @@ -402,7 +419,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -417,6 +434,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -427,7 +456,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -442,6 +471,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.expected.json b/snapshots/session/ralph-loop/tool-schemas.expected.json index deabaf4a0d..44c18da093 100644 --- a/snapshots/session/ralph-loop/tool-schemas.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "ralph", "description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.", @@ -359,7 +376,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -374,6 +391,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -384,7 +413,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -399,6 +428,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/session-query-spill/tool-schemas.expected.json b/snapshots/session/session-query-spill/tool-schemas.expected.json index 9d66381297..dd5544d7d4 100644 --- a/snapshots/session/session-query-spill/tool-schemas.expected.json +++ b/snapshots/session/session-query-spill/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -543,7 +560,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -558,6 +575,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -568,7 +597,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -583,6 +612,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json index 46abe32771..f61fc21202 100644 --- a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -389,7 +418,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -404,6 +433,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json index b8dc7c20a8..cd84466ebf 100644 --- a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json +++ b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json @@ -323,6 +323,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -402,7 +419,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -417,6 +434,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -427,7 +456,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -442,6 +471,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json index e48983205c..960e60afd0 100644 --- a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json @@ -174,6 +174,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -232,7 +249,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -247,6 +264,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -257,7 +286,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -272,6 +301,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json index d217c78441..3e450f980e 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -354,7 +371,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -369,6 +386,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -379,7 +408,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -394,6 +423,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json index 25793c1cd4..8ebad49ed3 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json index 45ab57c7c0..c38ea849b9 100644 --- a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json @@ -174,6 +174,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -253,7 +270,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -268,6 +285,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -278,7 +307,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -293,6 +322,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json +++ b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/team-targets/tool-schemas.expected.json b/snapshots/session/team-targets/tool-schemas.expected.json index 4926edb120..3409e4753a 100644 --- a/snapshots/session/team-targets/tool-schemas.expected.json +++ b/snapshots/session/team-targets/tool-schemas.expected.json @@ -251,6 +251,23 @@ "properties": {} } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", diff --git a/snapshots/session/text-turn/tool-schemas.expected.json b/snapshots/session/text-turn/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/text-turn/tool-schemas.expected.json +++ b/snapshots/session/text-turn/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/web-fetch/tool-schemas.expected.json b/snapshots/session/web-fetch/tool-schemas.expected.json index 5ef902c18a..11dba4b997 100644 --- a/snapshots/session/web-fetch/tool-schemas.expected.json +++ b/snapshots/session/web-fetch/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/workflow-confinement/tool-schemas.expected.json b/snapshots/session/workflow-confinement/tool-schemas.expected.json index 2b4b1f069f..2dd1f341df 100644 --- a/snapshots/session/workflow-confinement/tool-schemas.expected.json +++ b/snapshots/session/workflow-confinement/tool-schemas.expected.json @@ -260,6 +260,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -339,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -354,6 +371,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -364,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -379,6 +408,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/workspace-dependencies/tool-schemas.expected.json b/snapshots/session/workspace-dependencies/tool-schemas.expected.json index 4d3db178b9..cca2b4d36a 100644 --- a/snapshots/session/workspace-dependencies/tool-schemas.expected.json +++ b/snapshots/session/workspace-dependencies/tool-schemas.expected.json @@ -215,6 +215,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "load_workspace_dependencies", "description": "Get absolute paths to bundled Python and library directories, plus bundled Python distribution versions. Node.js and pnpm paths are included when the payload provides them. Python includes numpy, pandas, python-docx, python-pptx, openpyxl, Pillow, lxml, and XlsxWriter. Use these libraries for Office files unless the user or workspace instructions select another environment. When Node.js and pnpm paths are returned, run pnpm with that Node executable and pnpm script path. This does not change PATH or package-manager settings.", @@ -302,7 +319,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -317,6 +334,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -327,7 +356,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -342,6 +371,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/worktree/tool-schemas.expected.json b/snapshots/session/worktree/tool-schemas.expected.json index 6641f8f0b2..a18f81001c 100644 --- a/snapshots/session/worktree/tool-schemas.expected.json +++ b/snapshots/session/worktree/tool-schemas.expected.json @@ -277,6 +277,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +373,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -371,6 +388,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -381,7 +410,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -396,6 +425,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/tsconfig.base.json b/tsconfig.base.json index 86f3ef50fd..0a540ba263 100644 --- a/tsconfig.base.json +++ b/tsconfig.base.json @@ -185,6 +185,7 @@ "@deepseek-ai/dsh-tools/presentation": ["./packages/core/tools/src/presentation.ts"], "@deepseek-ai/dsh-tools/types": ["./packages/core/tools/src/types.ts"], "@deepseek-ai/dsh-tool-subagent-control/list-agents": ["./packages/subagent/tool-subagent-control/src/list-agents.ts"], + "@deepseek-ai/dsh-tool-subagent/list-models": ["./packages/subagent/tool-subagent/src/list-models.ts"], "@deepseek-ai/dsh-tool-subagent/model-selection-settings": ["./packages/subagent/tool-subagent/src/model-selection-settings.ts"], "@deepseek-ai/dsh-user-approval/types": ["./packages/interaction/user-approval/src/types.ts"], "@deepseek-ai/dsh-user-questions/types": ["./packages/interaction/user-questions/src/types.ts"], From 11c9d75ca34dceffd3928cde0cafad44836b57d5 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Sun, 27 Sep 2026 20:02:13 +0800 Subject: [PATCH 002/162] test(subagent): verify policy capture failure rolls back creation --- .../agent-loop/tests/serial-listener-review.spec.ts | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/packages/core/agent-loop/tests/serial-listener-review.spec.ts b/packages/core/agent-loop/tests/serial-listener-review.spec.ts index 5246d2cab4..032c286840 100644 --- a/packages/core/agent-loop/tests/serial-listener-review.spec.ts +++ b/packages/core/agent-loop/tests/serial-listener-review.spec.ts @@ -8,7 +8,6 @@ import AgentLoop from '@deepseek-ai/dsh-agent-loop' import JsonlSessionPersistence from '@deepseek-ai/dsh-session-persistence-jsonl' import { mountAgentLoopTestDependencies } from '@deepseek-ai/dsh-agent-loop-testkit' import { SessionId } from '@deepseek-ai/dsh-session' -import { defineContentToolFixture } from '@deepseek-ai/dsh-tools' import { bindScopeParent, createScope, scopeOf } from '@deepseek-ai/dsh-scope' import LocalFileReference from '@deepseek-ai/dsh-file-reference-local' import SubagentRuntime from '@deepseek-ai/dsh-subagent' @@ -99,11 +98,9 @@ describe('serial creation listener integrations', () => { } }) - it('rolls back creation when a shared preset tool installation fails', async () => { + it('rolls back creation when the shared preset cannot capture model-selection policy', async () => { const ctx = await core() - const errors: string[] = [] - ctx.logger.error = ((error: unknown) => { errors.push(String(error)) }) as typeof ctx.logger.error - await ctx.plugin(Selection) + await ctx.plugin(Selection, { enabled: true, allowedModels: [] }) await ctx.plugin(SubagentRuntime) await ctx.plugin(SubagentSpawn, { providerName: 'spawn' }) const preset = createScope(ctx, { preset: 'review' }) @@ -113,10 +110,8 @@ describe('serial creation listener integrations', () => { sessionId: SessionId('review-tool'), setup(agentCtx) { bindScopeParent(scopeOf(agentCtx)!, scopeOf(preset.ctx)!) - agentCtx.tools.register(defineContentToolFixture({ name: 'subagent', description: 'occupied', parameters: {}, execute: () => Promise.resolve([]) })) }, - })).rejects.toThrow('subagent') - expect(errors.some(error => error.includes('subagent'))).toBe(true) + })).rejects.toThrow('subagent model selection requires at least one allowed model') expect(ctx.agents.list()).toEqual([]) expect(ctx.sessions.list()).toEqual([]) } finally { From 7cf871d39aedf8c2bb1a742c9334ae372f586246 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Sun, 27 Sep 2026 20:12:16 +0800 Subject: [PATCH 003/162] test(subagent): refresh PowerShell and Python SDK tool snapshots --- .../dynamic-tools/tool-history.json | 3 ++ .../restart/requests.json | 2 + .../restart/session.1.v4.jsonl | 2 +- .../restart/session.2.v4.jsonl | 2 +- .../tool-schemas.expected.json | 45 ++++++++++++++++++- .../pwsh-tool-turn/tool-schemas.expected.json | 45 ++++++++++++++++++- 6 files changed, 93 insertions(+), 6 deletions(-) diff --git a/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json b/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json index 7a8a75aa85..d22dcd9dd6 100644 --- a/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json +++ b/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json @@ -56,6 +56,7 @@ "job_list", "job_output", "list_agents", + "list_subagent_models", "load_workspace_dependencies", "read", "read_image", @@ -113,6 +114,7 @@ "job_list", "job_output", "list_agents", + "list_subagent_models", "load_workspace_dependencies", "read", "read_image", @@ -185,6 +187,7 @@ "job_list", "job_output", "list_agents", + "list_subagent_models", "load_workspace_dependencies", "read", "read_image", diff --git a/scripts/snapshots/python-sdk-single-exe/restart/requests.json b/scripts/snapshots/python-sdk-single-exe/restart/requests.json index 06bbf97d4f..3c4c8b10d3 100644 --- a/scripts/snapshots/python-sdk-single-exe/restart/requests.json +++ b/scripts/snapshots/python-sdk-single-exe/restart/requests.json @@ -23,6 +23,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "subagent", "workflow", "working_directory" @@ -52,6 +53,7 @@ "job_kill", "job_list", "job_output", + "list_subagent_models", "subagent", "workflow", "working_directory" diff --git a/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl index 59d1880f27..3ce01a7a4d 100644 --- a/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{messageId}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Complete the first isolated Python SDK process turn."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Complete the first isolated Python","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{session-1}}","sessionFormatVersion":4,"throughSeq":12}} diff --git a/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl index be7dd18e39..cc12894292 100644 --- a/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{messageId}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Complete the second isolated Python SDK process turn."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Complete the second isolated Python","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{session-2}}","sessionFormatVersion":4,"throughSeq":12}} diff --git a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json index d1b4eef213..bee9fc050c 100644 --- a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json @@ -135,6 +135,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "pwsh", "description": "Run commands in a persistent PowerShell shell. State, including the current directory and exported environment variables, persists across calls for this agent.", @@ -214,7 +231,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -229,6 +246,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -239,7 +268,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -254,6 +283,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json index d1a1729417..dcf23f6b31 100644 --- a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json @@ -187,6 +187,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "pwsh", "description": "Execute a PowerShell command (`pwsh -Command`) and return its stdout/stderr. Each call runs in a fresh pwsh process; pass `workdir` instead of using `cd`. Paths use native Windows form (`C:\\...`); read environment variables with `$env:NAME`. Managed `$env:DSH_*` variables expose current harness environment facts. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. On Windows a force-killed command settles as `[exit code: 1]` without a signal marker — treat it as an interruption, not a command failure. Provide `description` before `command` in the arguments. Before any delete or move, verify that the resolved absolute target path is the intended one; never run it against a computed path you have not checked. Do not assign to automatic variables such as `$HOME`; variable names are case-insensitive, so `$home` is the same read-only variable. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under mode]`, a policy denial: do not retry another way. Under the Windows sandbox, read-only pwsh runs in PowerShell ConstrainedLanguage mode, while workspace-write stays in FullLanguage unless host policy says otherwise. In read-only, prefer cmdlets and core types (`[string]`, `[datetime]`, `[regex]`, `[guid]`); .NET static calls (`[System.IO.*]::`, `[math]::`), `Add-Type`, COM objects, and reflection fail with \"only core types\" errors. `-f` formatting, property access, and core cmdlets work. In both confined modes, programs cannot open named pipes, so a command that captures another program's output through piped stdio (Node.js `child_process.spawn`/`exec` with the default `stdio: 'pipe'`) fails with EPERM, while `stdio: 'inherit'` and `stdio: 'ignore'` spawns work and PowerShell's own pipelines are unaffected. That EPERM is the documented boundary: do not retry the command another way — escalate the exact command once or restructure it to avoid capturing output.", @@ -295,7 +312,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -310,6 +327,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -320,7 +349,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -335,6 +364,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ From 648eb70e906487c1a8cb46ed13c85ebbbc2d5e69 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Sun, 27 Sep 2026 20:19:44 +0800 Subject: [PATCH 004/162] test(subagent): refresh inherited parent tool list --- .../tests/expected/subagent-inheritance/parent.expected.jsonl | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl b/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl index 1cee326a29..01674e2a6e 100644 --- a/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl +++ b/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl @@ -17,8 +17,8 @@ {"type":"system/message","data":{"turn":2,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"sourceEventSeqs":[2],"surfaceOp":{"op":"replace","startSeq":2,"endSeq":2}} {"type":"user/message","data":{"content":[{"type":"text","text":"Delegate the write probe to a subagent."}],"source":{"kind":"user"},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: read-only. Any available operation enforced by the DSH file sandbox cannot modify files in the standing mode. Do not refuse a required modification from this policy alone: try an available tool normally and follow any denial and escalation guidance it returns.\n\nApproval policy: ask. Operations that require approval may ask through the configured answerers; without an available answerer, the request fails closed."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: read-only. Any available operation enforced by the DSH file sandbox cannot modify files in the standing mode. Do not refuse a required modification from this policy alone: try an available tool normally and follow any denial and escalation guidance it returns."},{"name":"approval:policy","text":"Approval policy: ask. Operations that require approval may ask through the configured answerers; without an available answerer, the request fails closed."}]},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","reasoningEffort":"low"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"resume","startsSeries":true}} -{"type":"developer/message","data":{"turn":2,"step":1,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":18},"surfaceOp":"append"} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","reasoningEffort":"low"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"resume","startsSeries":true}} +{"type":"developer/message","data":{"turn":2,"step":1,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"list_subagent_models"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":18},"surfaceOp":"append"} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Tighten this session to read-only.","messageSeqs":[3],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":2,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"delegate-write","name":"subagent","arguments":"{\"description\": \"Delegated write probe\", \"prompt\": \"Use the write tool exactly once with file_path set to exactly the relative path inherited.txt and content escaped. If the write is denied, reply with the single word CHILD_DENIED and the denial marker line; do not retry and do not request escalation. If it succeeds, reply CHILD_WROTE.\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"delegate-write","name":"subagent","args":["{\"description\": \"Delegated write probe\", \"prompt\": \"Use the write tool exactly once with file_path set to exactly the relative path inherited.txt and content escaped. If the write is denied, reply with the single word CHILD_DENIED and the denial marker line; do not retry and do not request escalation. If it succeeds, reply CHILD_WROTE.\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"delegate-write","name":"subagent","arguments":"{\"description\": \"Delegated write probe\", \"prompt\": \"Use the write tool exactly once with file_path set to exactly the relative path inherited.txt and content escaped. If the write is denied, reply with the single word CHILD_DENIED and the denial marker line; do not retry and do not request escalation. If it succeeds, reply CHILD_WROTE.\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} From 3cd55996176a57e580109cba48786badfde74494 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Sun, 27 Sep 2026 20:30:23 +0800 Subject: [PATCH 005/162] test(subagent): refresh Web model-selection tool snapshots --- .../plugin-instances.expected.md | 4 ++ apps/web/tests/shipped-composition.e2e.ts | 1 + .../tool-schemas.expected.json | 45 ++++++++++++++++++- .../web/ptc-round/system-prompt.expected.md | 24 +++++++++- 4 files changed, 70 insertions(+), 4 deletions(-) diff --git a/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md b/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md index b4584b0486..bafdc63177 100644 --- a/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md +++ b/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md @@ -3,6 +3,10 @@ - button "tool-subagent-control, 已启用": - strong: tool-subagent-control - text: Globally named send_message, interrupt_agent, and list_agents tools over ctx.subagents continuations + - listitem: + - button "tool-subagent/list-models, tool-subagent-list-models, 已启用": + - strong: tool-subagent/list-models + - code: tool-subagent-list-models - listitem: - button "tool-subagent-control/list-agents, tool-subagent-list-agents, 已启用": - strong: tool-subagent-control/list-agents diff --git a/apps/web/tests/shipped-composition.e2e.ts b/apps/web/tests/shipped-composition.e2e.ts index 6eefe81171..74e4a636b1 100644 --- a/apps/web/tests/shipped-composition.e2e.ts +++ b/apps/web/tests/shipped-composition.e2e.ts @@ -484,6 +484,7 @@ const EXPECTED_TOOLS = [ 'job_list', 'job_output', 'list_agents', + 'list_subagent_models', 'present', 'read', 'read_image', diff --git a/snapshots/web/fresh-round-trip/tool-schemas.expected.json b/snapshots/web/fresh-round-trip/tool-schemas.expected.json index 557f8adc57..02a74c4f52 100644 --- a/snapshots/web/fresh-round-trip/tool-schemas.expected.json +++ b/snapshots/web/fresh-round-trip/tool-schemas.expected.json @@ -323,6 +323,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "present", "description": "Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied.", @@ -702,7 +719,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -717,6 +734,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -727,7 +756,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -742,6 +771,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/web/ptc-round/system-prompt.expected.md b/snapshots/web/ptc-round/system-prompt.expected.md index 47cc71cb02..c5c077eb85 100644 --- a/snapshots/web/ptc-round/system-prompt.expected.md +++ b/snapshots/web/ptc-round/system-prompt.expected.md @@ -158,6 +158,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied. */ present: { /** Usually the 1-2 most important deliverables; at most 4 per call. */ @@ -281,7 +288,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -289,8 +296,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -298,6 +311,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -500,6 +519,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; present: { turn: number; files: { From 98d93db18f2cc8cc242b7a79fbb10f773c053f4e Mon Sep 17 00:00:00 2001 From: Yichen Jiang Date: Fri, 9 Oct 2026 13:38:12 +0800 Subject: [PATCH 006/162] feat(web): claim slash commands as atomic composer chips A claimed command is now a CommandChipNode on the draft's leading seat instead of styled literal text. The submit machine follows the chip: it holds the claim while the chip leads the document, releases it when the chip is deleted or displaced, and re-enters claimed when history restores the chip. Deleting at the chip's edge takes two steps: the first Backspace or Delete selects the chip, the next removes it. The claim-token decoration transform and the text-reference precedence rule are removed with the literal token. --- ...2026-10-09-composer-command-chip.i18n.yaml | 19 ++ .../2026-10-09-composer-command-chip.md | 40 ++++ .../2026-10-09-composer-command-chip.zh.md | 40 ++++ apps/web/tests/lifecycle-chrome.e2e.ts | 9 +- .../client/ui-conversation/README.i18n.yaml | 4 +- packages/client/ui-conversation/README.md | 4 +- packages/client/ui-conversation/README.zh.md | 4 +- .../src/client/contract/input.ts | 24 ++- .../src/client/input/editor/CommandChip.tsx | 28 +++ .../src/client/input/editor/claim-decor.ts | 73 ------- .../client/input/editor/command-chip-node.tsx | 149 +++++++++++++ .../input/editor/command-chip-selection.ts | 68 ++++++ .../input/editor/composer-editor.module.css | 5 + .../src/client/input/editor/keymap.ts | 4 +- .../src/client/input/editor/projection.ts | 8 +- .../src/client/input/editor/runtime.ts | 41 +++- .../src/client/input/editor/text-ref.ts | 16 +- .../src/client/input/facade.ts | 34 ++- .../src/client/input/machine.ts | 27 ++- .../src/client/skeleton/InputBar.module.css | 5 + .../tests/editor-focus.client.spec.ts | 2 +- .../tests/input-bar.client.spec.tsx | 19 +- .../tests/input-matrix.client.spec.tsx | 34 +-- .../tests/input-scenarios.client.spec.tsx | 199 ++++++++++++++---- .../tests/lexical-editor-core.client.spec.tsx | 140 ++++++++---- .../tests/submit-machine.client.spec.ts | 79 +++---- 26 files changed, 785 insertions(+), 290 deletions(-) create mode 100644 .agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml create mode 100644 .agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md create mode 100644 .agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md create mode 100644 packages/client/ui-conversation/src/client/input/editor/CommandChip.tsx delete mode 100644 packages/client/ui-conversation/src/client/input/editor/claim-decor.ts create mode 100644 packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx create mode 100644 packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts diff --git a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml new file mode 100644 index 0000000000..00e4d39595 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml @@ -0,0 +1,19 @@ +# Bilingual-pair consistency record for 2026-10-09-composer-command-chip.md (docs/i18n/README.md): per heading +# section, a hash of its English and Chinese blocks outside code blocks and generated regions. +# After editing either side, bring the other along and re-record with: +# pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md +/agent-note-claimed-commands-as-atomic-composer-chips: + en: 66d4f149698760c8 + zh: de219f394dfd4a4c +/agent-note-claimed-commands-as-atomic-composer-chips/problem: + en: 831c97e0de131639 + zh: ea35b4b179ff33f0 +/agent-note-claimed-commands-as-atomic-composer-chips/decision: + en: bde405a14ce2011e + zh: 272df3039ae260df +/agent-note-claimed-commands-as-atomic-composer-chips/alternatives-considered: + en: 9c694a21237d7772 + zh: b35ba7bf8b73b679 +/agent-note-claimed-commands-as-atomic-composer-chips/consequences: + en: ca28dd0d4e561d2e + zh: 0bb86fa41622bf09 diff --git a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md new file mode 100644 index 0000000000..366531ca5a --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md @@ -0,0 +1,40 @@ +# Agent Note: Claimed commands as atomic composer chips + +Status: implemented + +English | [中文](2026-10-09-composer-command-chip.zh.md) + +## Problem + +The [Lexical composer decision](../../archived/architecture/2026-08-20-web-composer-lexical-editor.md) kept a claimed command as literal leading text. A TextNode transform styled the token and split typed overflow back out of it, and the submit machine held the claim while the draft string started with the token. Every edit at the token boundary ran through that transform. Lexical merges same-style neighbours before the overflow node's own transform runs, so the split and the merge looped until Lexical aborted the update with error #14 and the DOM diverged from the editor state (issue #4490). PR #5480 cleared the copied style for the reported gesture; pasting an image after a claim and IME composition reach the same transform and remained unverified. The plain-text reference entity also needed a precedence rule to leave the claimed token alone. + +## Decision + +A claim replaces the leading trigger span with `CommandChipNode`, an atomic inline `DecoratorNode`, followed by the token's argument separator as an ordinary text node. + +- The node carries the `CommandClaim`, including its submit closure. History restores nodes with their fields, so undoing a chip deletion restores the claim. JSON export writes the token text; JSON import yields a plain text node because the closure has no JSON form. +- The detect projection counts the chip as one U+FFFC, so trigger detection never re-reads the command name. The clipboard projection is the token without its separator. The chip is not a reference occurrence. `EditorProjection.command` is the claim of a chip that opens the document. +- Outside a submit transaction, the machine's claim follows the document: `draft-changed` carries the leading chip's claim, so deleting or displacing the chip releases the claim and a history restore re-enters claimed. A failed command submission returns to claimed only when the draft is unchanged and the same chip still leads. A claim adjudicated from typed text at Enter has no chip and returns to plain on failure. +- A root transform keeps the chip on the leading seat. A chip with a preceding sibling or inside a later block becomes its literal token, which releases the claim as typing before the old token did. +- Deletion takes two steps. A character deletion that reaches the chip from a collapsed caret marks it selected (`data-selected` on its host; the root's `data-composer-chip-selected` hides the caret), and the next deletion in the same direction removes it whole. Any other key, a click, blur, or a caret that leaves the chip clears the mark. The editor selection stays a collapsed range and `isKeyboardSelectable()` stays false for the reason recorded on reference chips. Escape still does not release a claim. +- `claim-decor.ts` and the text-reference precedence rule are deleted. + +The `slash/input-*` events, `CommandClaim`, `InputState.claim`, and trigger sources keep their fields. `CommandClaim.token` still ends with the separator, and command arguments are still read from the clipboard text after it. + +## Alternatives considered + +**Keep literal text and repair each gesture.** Every gesture that reaches the boundary node needs its own proof, as #5480 showed, and the precedence rule with text references remains. + +**Reuse `ReferenceChipNode`.** Reference chips are occurrences serialized through their owner codec at submit; a command is not model content and must not enter `occurrences`. + +**Put the separator inside the chip text.** Lexical appends a managed line break after an inline decorator that ends its block, so the blank-arguments hint rendered by `p:last-child::after` would move to the next line. A separate text node keeps the hint on the chip's line. + +**Select the chip with a NodeSelection.** The plain-text binding's arrow, typing, and deletion handlers ignore non-range selections, so each of those gestures would need its own handling at the chip. + +**Keep claims in a shell map keyed by NodeKey.** The map duplicates the field the node already carries and needs pruning whenever history is cleared. + +## Consequences + +- Backspace no longer edits the command name character by character to reopen the menu. At the chip's edge, two Backspace or Delete presses remove the chip and leave the arguments as plain text. +- A draft restored after reload is plain text; Enter adjudication still submits it as a command. +- Whole-chip deletion through a real key event is observable only in the browser lane because jsdom lacks `Selection.modify`; `lifecycle-chrome.e2e.ts` pins the separator, select, and delete presses. diff --git a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md new file mode 100644 index 0000000000..1f9648a65b --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md @@ -0,0 +1,40 @@ +# Agent Note: 已认领命令作为输入框原子 chip + +Status: implemented + +[English](2026-10-09-composer-command-chip.md) | 中文 + +## 问题 + +[Lexical 输入框决策](../../archived/architecture/2026-08-20-web-composer-lexical-editor.md)把已认领命令保留为行首字面文本。一个 TextNode transform 给 token 加样式，并把输入进去的溢出文字拆出来；提交状态机在草稿字符串以 token 开头时保持认领。token 交界处的每次编辑都会经过这个 transform。Lexical 会在溢出节点自身的 transform 执行前合并样式相同的相邻节点，于是拆分与合并循环往复，直到 Lexical 以错误码 14 中止这次更新，DOM 与编辑器状态随之分叉（issue #4490）。PR #5480 针对报告的手势清除了被复制的样式；认领后粘贴图片和 IME 组字会进入同一个 transform，仍未验证。纯文本引用实体也需要一条优先级规则来避开已认领的 token。 + +## 决策 + +认领把行首触发区间替换为 `CommandChipNode`（一个原子 inline `DecoratorNode`），其后是作为普通文本节点的 token 参数分隔符。 + +- 节点携带 `CommandClaim`，包括其 submit 闭包。历史记录恢复节点时连同字段一起恢复，因此撤销 chip 删除会恢复认领。JSON 导出写入 token 文本；闭包没有 JSON 形式，因此 JSON 导入得到纯文本节点。 +- 检测投影把 chip 计为一个 U+FFFC，触发检测不会再读到命令名。剪贴板投影是不含分隔符的 token。chip 不是引用 occurrence。`EditorProjection.command` 是位于文档开头的 chip 所携带的认领。 +- 在提交事务之外，状态机的认领跟随文档：`draft-changed` 携带行首 chip 的认领，因此删除或挤开 chip 会释放认领，历史恢复会重新进入 claimed。命令提交失败时，只有草稿未变且同一个 chip 仍在行首才回到 claimed。Enter 时从输入文本裁决出的认领没有 chip，失败后回到 plain。 +- 一个根节点 transform 让 chip 保持在行首位置。前面有兄弟节点或位于后续块中的 chip 会变回字面 token，并像在旧 token 前输入那样释放认领。 +- 删除分两步。从折叠光标出发、方向触及 chip 的字符删除先把它标记为选中（宿主元素上的 `data-selected`；根元素的 `data-composer-chip-selected` 隐藏光标），同方向的下一次删除才整体移除它。其他按键、点击、失焦或光标离开 chip 都会清除标记。编辑器选区保持为折叠的 range，`isKeyboardSelectable()` 仍为 false，原因与引用 chip 记录的相同。Escape 仍不释放认领。 +- 删除 `claim-decor.ts` 和文本引用的优先级规则。 + +`slash/input-*` 事件、`CommandClaim`、`InputState.claim` 与触发源的字段保持不变。`CommandClaim.token` 仍以分隔符结尾，命令参数仍从其后的剪贴板文本读取。 + +## 考虑过的替代方案 + +**保留字面文本并逐个修复手势。** 如 #5480 所示，每个触及交界节点的手势都需要单独证明，与文本引用之间的优先级规则也仍然存在。 + +**复用 `ReferenceChipNode`。** 引用 chip 是在提交时经 owner codec 序列化的 occurrence；命令不是模型内容，不能进入 `occurrences`。 + +**把分隔符放进 chip 文本。** inline decorator 位于块末尾时，Lexical 会在其后追加受管换行，由 `p:last-child::after` 渲染的空参数提示会移到下一行。独立的文本节点让提示留在 chip 所在行。 + +**用 NodeSelection 选中 chip。** 纯文本绑定的方向键、输入和删除处理器都会忽略非 range 选区，这些手势在 chip 处都需要单独处理。 + +**在 shell 中用按 NodeKey 索引的表保存认领。** 该表重复了节点已携带的字段，并且每次清空历史时都需要修剪。 + +## 影响 + +- Backspace 不再逐字编辑命令名并重新打开菜单。在 chip 边缘按两次 Backspace 或 Delete 会删除 chip，参数保留为普通文字。 +- 刷新后恢复的草稿是纯文本；Enter 裁决仍会把它作为命令提交。 +- 通过真实按键整体删除 chip 只能在浏览器 lane 中观察，因为 jsdom 缺少 `Selection.modify`；`lifecycle-chrome.e2e.ts` 固定了删除分隔符、选中、删除这三次按键。 diff --git a/apps/web/tests/lifecycle-chrome.e2e.ts b/apps/web/tests/lifecycle-chrome.e2e.ts index a9226b7fb7..3a9ec7e07b 100644 --- a/apps/web/tests/lifecycle-chrome.e2e.ts +++ b/apps/web/tests/lifecycle-chrome.e2e.ts @@ -155,14 +155,14 @@ describe('web e2e: lifecycle & chrome (workspace flow / reload / dark mode)', () await inputPage.keyboard.insertText('这是任务') await expect.poll(() => input.textContent()).toBe(`${token} 这是任务`) for (let i = 0; i < 5; i++) await input.press('Backspace') - const tokenText = () => input.locator('[data-lexical-text][style*="business-primary"]').textContent() + const tokenText = () => input.locator('[data-composer-command]').textContent() await expect.poll(() => input.textContent()).toBe(token) await expect.poll(() => input.getAttribute('data-phase')).toBe('claimed') await expect.poll(tokenText).toBe(token) await input.press('Space') await expect.poll(() => input.getAttribute('data-phase')).toBe('claimed') await expect.poll(() => input.textContent()).toBe(`${token} `) - await expect.poll(async () => (await tokenText())?.trimEnd()).toBe(token) + await expect.poll(tokenText).toBe(token) const shownHint = () => input.locator('p').last().evaluate(element => getComputedStyle(element, '::after').content) await expect.poll(shownHint).toBe(JSON.stringify(hint)) const cdp = await inputPage.context().newCDPSession(inputPage) @@ -179,9 +179,14 @@ describe('web e2e: lifecycle & chrome (workspace flow / reload / dark mode)', () await expect.poll(shownHint).toBe('none') await cdp.send('Input.imeSetComposition', { text: '', selectionStart: 0, selectionEnd: 0 }) await expect.poll(shownHint).toBe(JSON.stringify(hint)) + // Backspace removes the separator, then selects the whole chip, then deletes it. await input.press('Backspace') await input.press('Backspace') + await expect.poll(() => input.locator('[data-composer-command][data-selected]').count()).toBe(1) + await expect.poll(() => input.getAttribute('data-phase')).toBe('claimed') + await input.press('Backspace') await expect.poll(() => input.getAttribute('data-phase')).toBe('plain') + await expect.poll(() => input.textContent()).toBe('') await writeComposerDraft(inputPage, input, '') const placeholder = inputPage.locator('[data-composer-placeholder]').first() await expect.poll(() => placeholder.isVisible()).toBe(true) diff --git a/packages/client/ui-conversation/README.i18n.yaml b/packages/client/ui-conversation/README.i18n.yaml index a22be7582c..e456dae371 100644 --- a/packages/client/ui-conversation/README.i18n.yaml +++ b/packages/client/ui-conversation/README.i18n.yaml @@ -18,8 +18,8 @@ en: 1d2a3a7f98c2cc9b zh: f895dd5e9f91e0a9 /deepseek-ai-dsh-client-ui-conversation/shell-and-standard-props: - en: 50b5e00a5fb7c2e5 - zh: 91333db3947f395d + en: 84cd1e10c9022adf + zh: 9fd87eeebe1d0f78 /deepseek-ai-dsh-client-ui-conversation/temporary-composer-entries: en: bb1244310f5c0382 zh: 824b84077432b559 diff --git a/packages/client/ui-conversation/README.md b/packages/client/ui-conversation/README.md index ec4221995f..207d38343b 100644 --- a/packages/client/ui-conversation/README.md +++ b/packages/client/ui-conversation/README.md @@ -66,7 +66,7 @@ The composer registers the File command action and owns its label, availability, Drafts retain plain text and file, folder, and Session reference chips across Session switching and reload. Public input methods accept text; internal restoration and Workspace transfer preserve structured references. The input model imports saved content before its first published read; React only mounts that editor and binds the existing Conversation store writer. The original storage key accepts legacy strings and writes semantic snapshots without runtime node IDs. A pending skill catalog does not block editing: slash names remain ordinary text until a matching catalog entry arrives. Arrival rematches the current text without changing chips, selection, or Undo history. Browser-owned attachment File objects are not persisted by this mechanism. The [draft initialization decision](../../../.agents/notes/implemented/architecture/2026-09-30-structured-draft-initialization.md) defines ownership and restoration. -Claimed commands retain their identity and highlight when only their arguments and trailing separator are deleted; editing the command name releases the claim. The same rules apply to every command and locale, including `/goal`, `/目标`, `/plan`, and `/计划`. Command hints and ordinary placeholders remain hidden throughout IME composition and reappear only after the editor commits the final text and the corresponding input is empty. +A claimed command is one atomic chip at the start of the draft, and the claim lasts exactly as long as that chip. Deleting the arguments and separator keeps the claim. At the chip's edge, the first Backspace or Delete selects the whole chip and the next one removes it and releases the claim, leaving the arguments as plain text; any other key, a click, or a caret move clears the selection. Text or a line break inserted before the chip turns it back into its literal token and releases the claim, and undo restores a deleted chip together with its claim. The same rules apply to every command and locale, including `/goal`, `/目标`, `/plan`, and `/计划`. Command hints and ordinary placeholders remain hidden throughout IME composition and reappear only after the editor commits the final text and the corresponding input is empty. Workspace selection uses `uiWorkspace.openWorkspace` to prepare the target and commit navigation. Draft content, including structured references, and attachments move in its synchronous preparation callback only while that request is current; later navigation or owner disposal leaves the original draft intact. @@ -82,7 +82,7 @@ Active transcripts in the main occurrence expose content-width drag handles in t The width-handle indicator follows the pointer only during a captured drag; ordinary hover leaves its position unchanged. -The resident composer survives no-Session and Session transitions. Whitespace hides its placeholder; a whitespace-only draft without attachments cannot be sent. The no-Session state keeps the same composer surface mounted but inert while the Workspace picker connects a blank Session. The surface is a shell-owned Lexical editor: reference chips are atomic decorator nodes carrying the owner's serialization identity (submission expands them through the owner codec), claimed slash commands stay styled leading text, folder text references carry the folder glyph as an icon prefix, and the draft's text and semantic references are persisted by the per-Session Conversation store. QueueDock reads `next-turn` from the Session `inbox` projection, including cold recovered messages, and excludes only requestIds still owned by local transcript submissions. Other queued rows retain their normal display and actions. Queue operations address exact queue occurrences through the scoped `ctx.conversation` service; queue previews render sent text through the shared inline reference projection from `ui-primitives` (wire session forms fold to their label) and show local or durable images and files in original attachment order. Images use thumbnails; files use compact name-and-size cards. An edit exposes the literal sent text in the ui-primitives `InlineEditor`, so line breaks survive re-editing. Durable thumbnails resolve through the session image URL cache. Busy Enter behavior is stored in the Host-backed `ui-conversation` settings namespace. The composer keymap arbitrates the trigger menu's keys through the slash pipeline — Tab settles the highlighted completion (or drills a drillable one), Escape and Shift+Tab leave the menu without settling — and leaves every other key to the editor. An overlay that takes the keyboard hands it back through `SessionInput.focus()`, which rides Lexical's own focus so the caret returns where the draft left it rather than at the start. +The resident composer survives no-Session and Session transitions. Whitespace hides its placeholder; a whitespace-only draft without attachments cannot be sent. The no-Session state keeps the same composer surface mounted but inert while the Workspace picker connects a blank Session. The surface is a shell-owned Lexical editor: reference chips are atomic decorator nodes carrying the owner's serialization identity (submission expands them through the owner codec), the claimed slash command is an atomic leading chip that carries its claim, folder text references carry the folder glyph as an icon prefix, and the draft's text and semantic references are persisted by the per-Session Conversation store. QueueDock reads `next-turn` from the Session `inbox` projection, including cold recovered messages, and excludes only requestIds still owned by local transcript submissions. Other queued rows retain their normal display and actions. Queue operations address exact queue occurrences through the scoped `ctx.conversation` service; queue previews render sent text through the shared inline reference projection from `ui-primitives` (wire session forms fold to their label) and show local or durable images and files in original attachment order. Images use thumbnails; files use compact name-and-size cards. An edit exposes the literal sent text in the ui-primitives `InlineEditor`, so line breaks survive re-editing. Durable thumbnails resolve through the session image URL cache. Busy Enter behavior is stored in the Host-backed `ui-conversation` settings namespace. The composer keymap arbitrates the trigger menu's keys through the slash pipeline — Tab settles the highlighted completion (or drills a drillable one), Escape and Shift+Tab leave the menu without settling — and leaves every other key to the editor. An overlay that takes the keyboard hands it back through `SessionInput.focus()`, which rides Lexical's own focus so the caret returns where the draft left it rather than at the start. Plain Enter uses the configured delivery mode, exactly Ctrl+Enter or Cmd+Enter uses its complement, and Shift+Enter inserts a line break. Enter with Alt, AltGraph, both Ctrl and Cmd, or Shift plus Ctrl/Cmd leaves the draft and command menu unchanged and keeps the DOM event available to application shortcuts. The Conversation plugin registers the fixed send, newline, complementary-delivery, command-menu, and reference-menu rows and reserves their bindings until it unloads. diff --git a/packages/client/ui-conversation/README.zh.md b/packages/client/ui-conversation/README.zh.md index bb0c423a36..806ad8c74c 100644 --- a/packages/client/ui-conversation/README.zh.md +++ b/packages/client/ui-conversation/README.zh.md @@ -66,7 +66,7 @@ target package 通过 declaration merge 扩展 snapshot 与 Location data map， 草稿中的普通文字及文件、目录、Session 引用胶囊在切换 Session 和刷新后保留。公开输入方法接受文本；内部恢复和工作区携带保留结构化引用。输入模型在首次对外读取前导入保存内容；React 只挂载同一编辑器并绑定现有 Conversation store 写入回调。原存储 key 兼容旧字符串，新写入保存不含运行时节点 ID 的语义快照。等待 skill 目录不会阻塞编辑：斜杠名称保持普通文字，直到匹配的目录项到达。目录到达只重匹配当前文字，不改变胶囊、选区或 Undo 历史。浏览器持有的附件 File 对象不由这一机制持久化。[草稿初始化决策](../../../.agents/notes/implemented/architecture/2026-09-30-structured-draft-initialization.zh.md)规定所有权与恢复流程。 -已认领的命令在仅删除参数和末尾分隔空格时保留身份与高亮，改动命令名才会释放认领。所有命令和语言使用相同规则，包括 `/goal`、`/目标`、`/plan` 和 `/计划`。输入法组合输入期间，命令提示和普通占位文字持续隐藏，直到编辑器提交最终文字且对应输入为空时才重新显示。 +已认领的命令是草稿开头的一个原子 chip，认领与该 chip 同生共灭。删除参数和分隔空格不影响认领。光标到达 chip 边缘时，第一次 Backspace 或 Delete 选中整个 chip，再按一次才删除它并释放认领，参数保留为普通文字；按其他键、点击或移动光标会取消选中。在 chip 前插入文字或换行会把它还原为字面 token 并释放认领，撤销会连同认领一起恢复被删除的 chip。所有命令和语言使用相同规则，包括 `/goal`、`/目标`、`/plan` 和 `/计划`。输入法组合输入期间，命令提示和普通占位文字持续隐藏，直到编辑器提交最终文字且对应输入为空时才重新显示。 工作区选择使用 `uiWorkspace.openWorkspace` 准备目标并提交导航。草稿内容（包括结构化引用）和附件仅在该请求仍为当前请求时，通过它的同步准备回调搬移；后续导航或所有者释放会保留原草稿。 @@ -82,7 +82,7 @@ Session 首次绑定或缓存的 Session 成为 current 时，shell 会在渲染 宽度拖拽条的指示线只在已捕获指针的拖拽期间跟随指针，普通悬停不改变其位置。 -常驻 composer 在无 Session 与有 Session 之间保持挂载。输入空白字符会隐藏占位提示；没有附件的纯空白草稿无法发送。无 Session 时，同一个编辑器表面保持 inert，Workspace picker 连接 blank Session。该表面是 shell 所有的 Lexical 编辑器：引用 chip 是携带 owner 序列化身份的原子 decorator 节点（提交时经 owner codec 展开），已认领的 slash command 保持为带样式的行首文本，文件夹文本引用以图标前缀携带文件夹图形，草稿文字和引用语义字段由逐 Session Conversation store 持久化。QueueDock 从 Session 的 `inbox` 投影读取 `next-turn`，包含从冷状态恢复的消息，仅排除仍由本地 transcript 提交承接的 requestId。其他排队行保留正常展示和操作。Queue 操作通过 scoped `ctx.conversation` service 寻址准确的 queue occurrence；queue 预览经 `ui-primitives` 的共享行内引用投影渲染已发送文本（wire 会话形式折叠为其标签），并按原始附件顺序展示本地或持久化的图片和文件。图片使用缩略图，文件使用紧凑的名称与大小卡片。编辑态在 ui-primitives 的 `InlineEditor` 中展示字面发送文本，因此重新编辑不会丢失换行。持久化缩略图通过会话图片 URL 缓存解析。繁忙时 Enter 行为保存在 Host-backed `ui-conversation` settings namespace。 composer 键盘映射经斜杠流水线裁决触发菜单的按键——Tab 确认高亮补全项（可下钻项则下钻），Escape 与 Shift+Tab 离开菜单且不选定——其余按键交给编辑器自身。 接管键盘的浮层通过 `SessionInput.focus()` 把键盘还回来，该路径走 Lexical 自己的 focus，因此光标回到草稿原来的位置而不是开头。 +常驻 composer 在无 Session 与有 Session 之间保持挂载。输入空白字符会隐藏占位提示；没有附件的纯空白草稿无法发送。无 Session 时，同一个编辑器表面保持 inert，Workspace picker 连接 blank Session。该表面是 shell 所有的 Lexical 编辑器：引用 chip 是携带 owner 序列化身份的原子 decorator 节点（提交时经 owner codec 展开），已认领的 slash command 是携带认领的行首原子 chip，文件夹文本引用以图标前缀携带文件夹图形，草稿文字和引用语义字段由逐 Session Conversation store 持久化。QueueDock 从 Session 的 `inbox` 投影读取 `next-turn`，包含从冷状态恢复的消息，仅排除仍由本地 transcript 提交承接的 requestId。其他排队行保留正常展示和操作。Queue 操作通过 scoped `ctx.conversation` service 寻址准确的 queue occurrence；queue 预览经 `ui-primitives` 的共享行内引用投影渲染已发送文本（wire 会话形式折叠为其标签），并按原始附件顺序展示本地或持久化的图片和文件。图片使用缩略图，文件使用紧凑的名称与大小卡片。编辑态在 ui-primitives 的 `InlineEditor` 中展示字面发送文本，因此重新编辑不会丢失换行。持久化缩略图通过会话图片 URL 缓存解析。繁忙时 Enter 行为保存在 Host-backed `ui-conversation` settings namespace。 composer 键盘映射经斜杠流水线裁决触发菜单的按键——Tab 确认高亮补全项（可下钻项则下钻），Escape 与 Shift+Tab 离开菜单且不选定——其余按键交给编辑器自身。 接管键盘的浮层通过 `SessionInput.focus()` 把键盘还回来，该路径走 Lexical 自己的 focus，因此光标回到草稿原来的位置而不是开头。 普通 Enter 使用已配置的投递模式，严格的 Ctrl+Enter 或 Cmd+Enter 使用互补模式，Shift+Enter 插入换行。带 Alt、AltGraph、同时带 Ctrl 与 Cmd，或 Shift 与 Ctrl/Cmd 的 Enter 保持草稿和指令菜单不变，并将 DOM 事件留给应用快捷键。Conversation 插件注册发送、换行、互补投递、命令菜单和引用菜单的固定条目，并在卸载前保留这些键位。 diff --git a/packages/client/ui-conversation/src/client/contract/input.ts b/packages/client/ui-conversation/src/client/contract/input.ts index 52ec80eda9..416e46b463 100644 --- a/packages/client/ui-conversation/src/client/contract/input.ts +++ b/packages/client/ui-conversation/src/client/contract/input.ts @@ -39,7 +39,7 @@ export interface SubmitOutcome { export interface CommandClaim { /** Catalog command name without the leading slash (the key of per-command composer copy such as `hint.*`). */ readonly name: string - /** Inserted command text with its argument separator; the bare complete name also retains the claim. */ + /** Command spelling followed by its argument separator; the composer shows the spelling as a chip and inserts the separator after it. */ readonly token: string readonly hint?: string readonly attachments?: boolean @@ -167,7 +167,7 @@ export type DraftAttachmentId = Branded<'DraftAttachmentId'> * applied the edit after phase and span guards). */ export interface InputTarget { - /** Replace the trigger span with claim.token and enter claimed (span-CAS'd). */ + /** Replace the leading trigger span with a command chip and enter claimed (span-CAS'd). */ beginCommand(claim: CommandClaim, span: TokenSpan): boolean /** Replace the trigger span with one reference chip (span-CAS'd). */ insertReference(ref: ReferenceInsert, span: TokenSpan): boolean @@ -310,19 +310,27 @@ export interface SubmitAttempt { /** * Submit-machine input events (the machine's single write path). Text * mutation lives in the editor; the machine only observes the draft through - * event payloads (claim integrity, enter snapshots, settlement decisions). + * event payloads (leading command chip, enter snapshots, settlement decisions). */ export type InputEvent = - /** Clipboard projection changed: the claimed integrity watch runs (zero effects). */ - | { readonly type: 'draft-changed'; readonly draft: string } - /** The editor applied a claim-token replacement: enter claimed. */ + /** Editor content changed; carries the claim of the leading command chip, if any (zero effects). */ + | { readonly type: 'draft-changed'; readonly command: CommandClaim | undefined } + /** The editor inserted a command chip: enter claimed. */ | { readonly type: 'claim'; readonly claim: CommandClaim } /** Enter submission with the current clipboard projection. */ | { readonly type: 'enter'; readonly mode: InputSubmitMode; readonly draft: string; readonly submission?: MessageSubmission } | { readonly type: 'adjudicated'; readonly attempt: SubmitAttempt; readonly outcome: PickOutcome } | { readonly type: 'adjudication-failed'; readonly attempt: SubmitAttempt; readonly message: string } - /** Settlement carries the live clipboard projection for suffix-retention and claim re-entry decisions. */ - | { readonly type: 'submit-settled'; readonly attempt: SubmitAttempt; readonly ok: boolean; readonly draft: string; readonly outcome?: SubmitOutcome; readonly message?: string } + /** Settlement carries the live clipboard projection and leading command chip claim for claim re-entry decisions. */ + | { + readonly type: 'submit-settled' + readonly attempt: SubmitAttempt + readonly ok: boolean + readonly draft: string + readonly command: CommandClaim | undefined + readonly outcome?: SubmitOutcome + readonly message?: string + } /** Settlement of one optimistic default send, independent of the frozen command slot. */ | { readonly type: 'sink-settled'; readonly attempt: SubmitAttempt; readonly ok: boolean; readonly outcome?: SubmitOutcome; readonly message?: string } /** Commit an attachment-only send whose empty draft did not need an attempt. */ diff --git a/packages/client/ui-conversation/src/client/input/editor/CommandChip.tsx b/packages/client/ui-conversation/src/client/input/editor/CommandChip.tsx new file mode 100644 index 0000000000..f4ef8f7658 --- /dev/null +++ b/packages/client/ui-conversation/src/client/input/editor/CommandChip.tsx @@ -0,0 +1,28 @@ +/** + * Visual body of the claimed command chip: the CommandChipNode's React face. + * Pure display with the reference chip's treatment; the claim and its + * lifecycle live on the node and the shell. + */ +import clsx from 'clsx' +import type { ReactNode } from 'react' +import css from './ReferenceChip.module.css' +import referenceCss from './composer-editor.module.css' + +/** Display inputs of the command chip. */ +export interface CommandChipProps { + /** Claimed token without its argument separator, e.g. `/goal`. */ + readonly token: string +} + +/** + * Render the claimed command chip. + * @param props - the token to display. + * @returns the chip body. + */ +export function CommandChip({ token }: CommandChipProps): ReactNode { + return ( + + {token} + + ) +} diff --git a/packages/client/ui-conversation/src/client/input/editor/claim-decor.ts b/packages/client/ui-conversation/src/client/input/editor/claim-decor.ts deleted file mode 100644 index e0ed9d3e57..0000000000 --- a/packages/client/ui-conversation/src/client/input/editor/claim-decor.ts +++ /dev/null @@ -1,73 +0,0 @@ -/** - * Claim-token highlight: while a command claim holds, the draft's leading - * token renders in the business accent. A TextNode transform keeps the token - * in its own styled node (splitting when typing merges text into it), and the - * shell nudges the first leaf dirty when the claim flips so entering and - * leaving claimed restyles without a text edit. - */ -import type { LexicalEditor, TextNode as TextNodeType } from 'lexical' -import { $getRoot, $isElementNode, $isTextNode, TextNode } from 'lexical' - -/** Inline style carried by the claim-token node. */ -const TOKEN_STYLE = 'color: var(--dsw-alias-state-business-primary)' - -/** The document's first text leaf, or null (empty document / leading chip). */ -function firstTextLeaf(): TextNodeType | null { - const block = $getRoot().getFirstChild() - if (!$isElementNode(block)) return null - const leaf = block.getFirstChild() - return $isTextNode(leaf) ? leaf : null -} - -/** - * Register the claim-token styling transform. - * @param editor - the shell-owned editor. - * @param activeToken - live claim token accessor; null while unclaimed. - * @returns the unregister disposer. - */ -export function registerClaimDecoration(editor: LexicalEditor, activeToken: () => string | null): () => void { - return editor.registerNodeTransform(TextNode, (node) => { - const first = firstTextLeaf() - if (first === null || node.getKey() !== first.getKey()) { - // Off the token seat: clear a stale token style (a node can move here - // by paragraph merges). - if (node.getStyle() === TOKEN_STYLE) node.setStyle('') - return - } - const text = node.getTextContent() - const active = activeToken() - const token = text === active?.trimEnd() ? text : active - if (token === null || !text.startsWith(token)) { - if (node.getStyle() === TOKEN_STYLE) node.setStyle('') - return - } - if (text.length > token.length) { - // Typing at the token boundary lands in the styled node; split the - // overflow back out so only the token itself carries the color. The - // split copies the token style onto the overflow; it is cleared in this - // pass because Lexical's normalization merges same-style neighbours - // before the overflow's own transform runs, which would rejoin the two - // halves and loop until Lexical aborts the update. - const [tokenNode, overflow] = node.splitText(token.length) - if (tokenNode !== undefined && tokenNode.getStyle() !== TOKEN_STYLE) tokenNode.setStyle(TOKEN_STYLE) - overflow?.setStyle('') - return - } - if (node.getStyle() !== TOKEN_STYLE) node.setStyle(TOKEN_STYLE) - }) -} - -/** - * Nudge the token seat dirty so the transform restyles after a claim flip - * (claims change phase without a text edit; transforms only run on dirty - * nodes). - * @param editor - the shell-owned editor. - */ -export function refreshClaimDecoration(editor: LexicalEditor): void { - // Not discrete: a refresh can fire from inside an update listener, where a - // synchronous nested commit would recurse; the queued update lands on the - // next flush. - editor.update(() => { - firstTextLeaf()?.markDirty() - }) -} diff --git a/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx b/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx new file mode 100644 index 0000000000..2b2760a36e --- /dev/null +++ b/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx @@ -0,0 +1,149 @@ +/** + * CommandChipNode: the claimed slash command as one atomic Lexical decorator + * on the document's leading seat. The node carries the claim it was inserted + * for, so a history restore of the node restores the claim with it. + * `getTextContent()` answers the token without its argument separator; the + * separator is an ordinary text node after the chip. The detect projection + * counts the chip as one U+FFFC; see projection.ts. + */ +import type { JSX } from 'react' +import type { + LexicalEditor, LexicalNode, NodeKey, SerializedLexicalNode, Spread, +} from 'lexical' +import { $createTextNode, DecoratorNode, RootNode } from 'lexical' +import { $dfs } from '@lexical/utils' +import type { CommandClaim } from '../../contract/input.ts' +import { CommandChip } from './CommandChip.tsx' + +/** JSON form of one command chip (Lexical node serialization contract). */ +export type SerializedCommandChipNode = Spread<{ text: string }, SerializedLexicalNode> + +/** One claimed command as an atomic decorator node. */ +export class CommandChipNode extends DecoratorNode { + /** The claim this chip anchors; its submit closure has no JSON form. */ + __claim: CommandClaim + + /** Lexical node registry type tag. */ + static override getType(): string { + return 'command-chip' + } + + /** + * Clone with identity (Lexical writable-copy contract). + * @param node - node to clone. + * @returns a copy carrying the same NodeKey and claim. + */ + static override clone(node: CommandChipNode): CommandChipNode { + return new CommandChipNode(node.__claim, node.__key) + } + + /** + * Rebuild one chip from its JSON form. JSON cannot carry the claim's + * submit closure, so an imported command is its literal token text. + * @param json - serialized chip. + * @returns a plain text node holding the token. + */ + static override importJSON(json: SerializedCommandChipNode): LexicalNode { + return $createTextNode(json.text) + } + + /** + * @param claim - the command claim the chip anchors. + * @param key - Lexical clone-path key; absent for fresh nodes. + */ + constructor(claim: CommandClaim, key?: NodeKey) { + super(key) + this.__claim = claim + } + + /** Serialize to the JSON node form (the token text alone). */ + override exportJSON(): SerializedCommandChipNode { + return { + ...super.exportJSON(), + type: 'command-chip', + version: 1, + text: this.getTextContent(), + } + } + + /** + * Mount the chip's host element; the decorator portal renders into it. + * @returns an inline, non-editable span carrying the test/e2e anchor. + */ + override createDOM(): HTMLElement { + const el = document.createElement('span') + el.setAttribute('data-composer-command', this.__claim.name) + el.setAttribute('contenteditable', 'false') + return el + } + + /** Host element never changes shape. */ + override updateDOM(): boolean { + return false + } + + /** The chip sits in the text line. */ + override isInline(): boolean { + return true + } + + /** + * Arrows step across the chip and Backspace/Delete remove it whole, for + * the reason documented on ReferenceChipNode.isKeyboardSelectable. + */ + override isKeyboardSelectable(): boolean { + return false + } + + /** Clipboard / persistence projection: the token without its separator. */ + override getTextContent(): string { + return this.__claim.token.trimEnd() + } + + /** The claim this chip anchors. */ + getClaim(): CommandClaim { + return this.getLatest().__claim + } + + /** React face rendered into the host element by the decorator portal. */ + override decorate(): JSX.Element { + return + } +} + +/** + * Mint one chip node for a claim. + * @param claim - the command claim. + * @returns the fresh node. + */ +export function $createCommandChipNode(claim: CommandClaim): CommandChipNode { + return new CommandChipNode(claim) +} + +/** + * Command chip type guard. + * @param node - any node or nullish. + * @returns whether the node is a CommandChipNode. + */ +export function $isCommandChipNode(node: LexicalNode | null | undefined): node is CommandChipNode { + return node instanceof CommandChipNode +} + +/** + * Keep command chips on the leading seat. A chip anywhere but the first + * child of the first block (text typed or a line break inserted before it) + * becomes its literal token, which releases the claim. The root transform + * runs whenever any attached node is dirty. + * @param editor - the shell-owned editor. + * @returns the unregister disposer. + */ +export function registerCommandChipSeat(editor: LexicalEditor): () => void { + return editor.registerNodeTransform(RootNode, (root) => { + for (const { node } of $dfs(root)) { + if ($isCommandChipNode(node) + && (node.getPreviousSibling() !== null || node.getParentOrThrow().getPreviousSibling() !== null)) { + node.replace($createTextNode(node.getTextContent())) + } + } + }) +} diff --git a/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts b/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts new file mode 100644 index 0000000000..a6c33f0597 --- /dev/null +++ b/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts @@ -0,0 +1,68 @@ +/** + * Two-step command chip deletion. A character deletion (Backspace, Delete, + * or an input-event deletion) whose direction reaches the command chip from a + * collapsed caret first selects the chip; the next deletion in that direction + * removes it. Any other key, a click, blur, or a caret that no longer touches + * the chip clears the selection. The selected state is presentation on the + * chip's host and the root element; the editor selection stays a collapsed + * range beside the chip, so typing and arrows keep their plain-text behavior. + */ +import type { LexicalEditor, NodeKey } from 'lexical' +import { + $getAdjacentNode, $getSelection, $isRangeSelection, BLUR_COMMAND, CLICK_COMMAND, COMMAND_PRIORITY_CRITICAL, + COMMAND_PRIORITY_HIGH, DELETE_CHARACTER_COMMAND, KEY_DOWN_COMMAND, +} from 'lexical' +import { mergeRegister } from '@lexical/utils' +import { $isCommandChipNode } from './command-chip-node.tsx' +import type { CommandChipNode } from './command-chip-node.tsx' + +/** Keys that keep a selected chip: deletions and bare modifiers. */ +const RETAINING_KEYS = new Set(['Backspace', 'Delete', 'Shift', 'Control', 'Alt', 'Meta']) + +/** The chip a collapsed caret deletes next in one direction, or null. */ +function $commandChipBesideCaret(isBackward: boolean): CommandChipNode | null { + const selection = $getSelection() + if (!$isRangeSelection(selection) || !selection.isCollapsed()) return null + const node = $getAdjacentNode(selection.focus, isBackward) + return $isCommandChipNode(node) ? node : null +} + +/** + * Register two-step deletion for the command chip. + * @param editor - the shell-owned editor. + * @returns the unregister disposer, which also clears a selected chip. + */ +export function registerCommandChipSelection(editor: LexicalEditor): () => void { + let selected: { readonly key: NodeKey; readonly isBackward: boolean } | null = null + const paint = (next: typeof selected): void => { + if (selected !== null && selected.key !== next?.key) { + editor.getElementByKey(selected.key)?.removeAttribute('data-selected') + } + selected = next + if (next !== null) editor.getElementByKey(next.key)?.setAttribute('data-selected', '') + editor.getRootElement()?.toggleAttribute('data-composer-chip-selected', next !== null) + } + const clear = (): false => { + if (selected !== null) paint(null) + return false + } + return mergeRegister( + editor.registerCommand(DELETE_CHARACTER_COMMAND, (isBackward) => { + const chip = $commandChipBesideCaret(isBackward) + if (chip === null || (selected?.key === chip.getKey() && selected.isBackward === isBackward)) return clear() + paint({ key: chip.getKey(), isBackward }) + return true + }, COMMAND_PRIORITY_HIGH), + editor.registerCommand(KEY_DOWN_COMMAND, event => RETAINING_KEYS.has(event.key) ? false : clear(), COMMAND_PRIORITY_CRITICAL), + editor.registerCommand(CLICK_COMMAND, clear, COMMAND_PRIORITY_CRITICAL), + editor.registerCommand(BLUR_COMMAND, clear, COMMAND_PRIORITY_CRITICAL), + editor.registerUpdateListener(({ editorState }) => { + const current = selected + if (current === null) return + const touching = editorState.read(() => $commandChipBesideCaret(current.isBackward)?.getKey() === current.key) + // A restore can remount the chip's host element; repaint keeps the mark on it. + paint(touching ? current : null) + }), + () => { paint(null) }, + ) +} diff --git a/packages/client/ui-conversation/src/client/input/editor/composer-editor.module.css b/packages/client/ui-conversation/src/client/input/editor/composer-editor.module.css index 0eefd223a1..3d153065b4 100644 --- a/packages/client/ui-conversation/src/client/input/editor/composer-editor.module.css +++ b/packages/client/ui-conversation/src/client/input/editor/composer-editor.module.css @@ -18,6 +18,11 @@ cursor: pointer; } +/* A command chip selected by its first deletion keystroke keeps the hover tint until the next deletion removes it. */ +[data-composer-command][data-selected] > .reference { + background: var(--dsw-alias-state-business-tertiary); +} + .textRef { display: inline-block; max-width: 100%; diff --git a/packages/client/ui-conversation/src/client/input/editor/keymap.ts b/packages/client/ui-conversation/src/client/input/editor/keymap.ts index e105a1d6c0..b4fe576e19 100644 --- a/packages/client/ui-conversation/src/client/input/editor/keymap.ts +++ b/packages/client/ui-conversation/src/client/input/editor/keymap.ts @@ -115,7 +115,7 @@ export function registerComposerKeymap(editor: LexicalEditor, handlers: Composer ), editor.registerCommand(KEY_ESCAPE_COMMAND, (event) => { // Escape layering: an open overlay closes; claimed without an overlay - // does NOT release (backspacing the token is the only exit gesture). + // does NOT release (deleting the command chip is the only exit gesture). handlers.dismissPopup() if (handlers.arbitrate('escape', isComposingEvent(event, recentlyComposing)) === 'consumed') { event.preventDefault() @@ -127,7 +127,7 @@ export function registerComposerKeymap(editor: LexicalEditor, handlers: Composer if (isComposingEvent(event, recentlyComposing)) return false const consumed = handlers.space() if (consumed) { - event.preventDefault() // claim token already carries the trailing separator + event.preventDefault() // the command chip insertion already carries its separator return true } return false diff --git a/packages/client/ui-conversation/src/client/input/editor/projection.ts b/packages/client/ui-conversation/src/client/input/editor/projection.ts index f48a5b595d..c48c97ad8a 100644 --- a/packages/client/ui-conversation/src/client/input/editor/projection.ts +++ b/packages/client/ui-conversation/src/client/input/editor/projection.ts @@ -12,7 +12,9 @@ import { $getRoot, $getSelection, $isElementNode, $isLineBreakNode, $isRangeSelection, $isTextNode, } from 'lexical' import type { Occurrence } from '../../contract/draft-editor.ts' +import type { CommandClaim } from '../../contract/input.ts' import { $isReferenceChipNode } from './chip-node.tsx' +import { $isCommandChipNode } from './command-chip-node.tsx' /** The detect-projection stand-in for one chip (object replacement character). */ export const ATOMIC_CHAR = '' @@ -79,7 +81,7 @@ export function $composerLayout(): ComposerLayout { const kids = element.getChildren() children.set(element.getKey(), kids.map(kid => kid.getKey())) for (const kid of kids) { - if ($isReferenceChipNode(kid)) { + if ($isReferenceChipNode(kid) || $isCommandChipNode(kid)) { pushLeaf('chip', kid, ATOMIC_CHAR, kid.getTextContent()) } else if ($isTextNode(kid)) { const text = kid.getTextContent() @@ -162,6 +164,8 @@ export interface EditorProjection { readonly selection: { readonly start: number; readonly end: number } | null /** Collapsed caret in detect coordinates; null while the selection is absent or ranged. */ readonly caret: number | null + /** Claim of the command chip opening the document; undefined without one. */ + readonly command: CommandClaim | undefined } /** @@ -211,6 +215,7 @@ export function $projectComposer(idOf: (key: NodeKey) => number): EditorProjecti ...(chip.isInvalid() ? { invalid: true } : {}), }) } + const lead = layout.segments[0]?.node const selection = $getSelection() let range: { start: number; end: number } | null = null if ($isRangeSelection(selection)) { @@ -226,5 +231,6 @@ export function $projectComposer(idOf: (key: NodeKey) => number): EditorProjecti occurrences, selection: range, caret: range !== null && range.start === range.end ? range.start : null, + command: $isCommandChipNode(lead) ? lead.getClaim() : undefined, } } diff --git a/packages/client/ui-conversation/src/client/input/editor/runtime.ts b/packages/client/ui-conversation/src/client/input/editor/runtime.ts index 7cbac8676b..1da9a1c10e 100644 --- a/packages/client/ui-conversation/src/client/input/editor/runtime.ts +++ b/packages/client/ui-conversation/src/client/input/editor/runtime.ts @@ -10,9 +10,11 @@ import { registerPlainText } from '@lexical/plain-text' import { createEmptyHistoryState, registerHistory } from '@lexical/history' import { mergeRegister } from '@lexical/utils' import type { DraftReference, ReferenceInsert } from '../../contract/draft-editor.ts' +import type { CommandClaim } from '../../contract/input.ts' import { registerReferenceActivation } from './reference-activation.ts' import { ReferenceChipNode, $createReferenceChipNode } from './chip-node.tsx' -import { refreshClaimDecoration, registerClaimDecoration } from './claim-decor.ts' +import { CommandChipNode, $createCommandChipNode, registerCommandChipSeat } from './command-chip-node.tsx' +import { registerCommandChipSelection } from './command-chip-selection.ts' import { registerTextRefDecoration, rescanTextRefs, TextRefNode } from './text-ref.ts' import type { EditorProjection } from './projection.ts' import { $composerLayout, $projectComposer, detectOffsetOfClipboardOffset } from './projection.ts' @@ -25,7 +27,6 @@ type Lexicon = ReadonlyMap<'/' | '@', readonly string[]> interface DraftEditorRuntimeDeps { readonly onUpdate: () => void readonly openReference: (source: string | undefined, reference: Pick) => boolean - readonly activeClaimToken: () => string | null readonly lexicon: () => Lexicon readonly resolveLexicon: () => ObservableSnapshot | undefined } @@ -45,7 +46,9 @@ const HISTORY_MERGE_DELAY_MS = 1000 export class DraftEditorRuntime { /** The editor bound by the Composer's contenteditable host. */ readonly editor: LexicalEditor - private projected: EditorProjection = { detectText: '', clipboardText: '', occurrences: [], selection: null, caret: null } + private projected: EditorProjection = { + detectText: '', clipboardText: '', occurrences: [], selection: null, caret: null, command: undefined, + } /** Stable occurrence ids per chip NodeKey (undo restores keys, so ids survive it too). */ private readonly occurrenceIds = new Map() private occurrenceSeq = 0 @@ -57,7 +60,7 @@ export class DraftEditorRuntime { constructor(private readonly deps: DraftEditorRuntimeDeps) { this.editor = createEditor({ namespace: 'dsh-composer', - nodes: [ReferenceChipNode, TextRefNode], + nodes: [ReferenceChipNode, CommandChipNode, TextRefNode], onError: (error) => { throw error }, }) } @@ -88,8 +91,9 @@ export class DraftEditorRuntime { this.deps.openReference(source, reference)), registerHistory(this.editor, createEmptyHistoryState(), HISTORY_MERGE_DELAY_MS), this.editor.registerUpdateListener(() => { this.deps.onUpdate() }), - registerClaimDecoration(this.editor, () => this.deps.activeClaimToken()), - registerTextRefDecoration(this.editor, () => this.deps.lexicon(), () => this.deps.activeClaimToken()), + registerCommandChipSeat(this.editor), + registerCommandChipSelection(this.editor), + registerTextRefDecoration(this.editor, () => this.deps.lexicon()), () => { this.lexiconOff?.() this.lexiconOff = undefined @@ -261,6 +265,26 @@ export class DraftEditorRuntime { return applied } + /** + * Replace a span with a command chip followed by the claim token's + * argument separator, unless the separator already follows. + * @param span - detect-coordinate range. + * @param claim - the command claim the chip anchors. + * @param tail - the character following the range before editing. + * @returns whether the range mapped and the edit applied. + */ + insertCommand(span: DetectSpan, claim: CommandClaim, tail: string): boolean { + const separator = claim.token.slice(claim.token.trimEnd().length) + let applied = false + this.applyEdit(() => { + const nodes = separator === '' || tail === separator.charAt(0) + ? [$createCommandChipNode(claim)] + : [$createCommandChipNode(claim), $createTextNode(separator)] + applied = $replaceDetectSpanWithNodes(span, nodes) + }) + return applied + } + /** * Insert an ordered file-reference batch after the live selection without deleting it. * @param references - validated references in source order. @@ -280,11 +304,6 @@ export class DraftEditorRuntime { return applied } - /** Refresh claim-token decoration after the model's claim changes. */ - refreshClaimDecoration(): void { - refreshClaimDecoration(this.editor) - } - /** * Clear committed content using the model's suffix decision inside the editor update. * @param prefixLength - returns the clipboard-prefix length to remove, or null to clear the root. diff --git a/packages/client/ui-conversation/src/client/input/editor/text-ref.ts b/packages/client/ui-conversation/src/client/input/editor/text-ref.ts index dc0dccc74b..5221b38a57 100644 --- a/packages/client/ui-conversation/src/client/input/editor/text-ref.ts +++ b/packages/client/ui-conversation/src/client/input/editor/text-ref.ts @@ -70,28 +70,18 @@ export class TextRefNode extends TextNode { } /** - * Register the plain-text reference entity transform. The claim decoration - * has precedence on the leading-token seat: while a command claim holds, the - * claimed token must stay a plain TextNode (transforms register per concrete - * node class, so a TextRefNode would never receive the TextNode claim - * transform and the warn color would be lost). + * Register the plain-text reference entity transform. * @param editor - the shell-owned editor. * @param lexiconOf - live per-trigger name-roll accessor (the controller's aggregated store). - * @param activeToken - live claim token accessor; null while unclaimed. * @returns the unregister disposer. */ export function registerTextRefDecoration( editor: LexicalEditor, lexiconOf: () => ReadonlyMap<'/' | '@', readonly string[]>, - activeToken: () => string | null, ): () => void { const getMatch = (text: string): { start: number; end: number } | null => { - const claim = activeToken() - for (const range of scanTextRefs(text, lexiconOf())) { - if (claim !== null && range.start === 0 && text.slice(range.start, range.end) === claim.trimEnd()) continue - return { start: range.start, end: range.end } - } - return null + const range = scanTextRefs(text, lexiconOf())[0] + return range === undefined ? null : { start: range.start, end: range.end } } return mergeRegister( ...registerLexicalTextEntity( diff --git a/packages/client/ui-conversation/src/client/input/facade.ts b/packages/client/ui-conversation/src/client/input/facade.ts index 9f4a264e46..b3d7d74fd0 100644 --- a/packages/client/ui-conversation/src/client/input/facade.ts +++ b/packages/client/ui-conversation/src/client/input/facade.ts @@ -86,6 +86,7 @@ function guardOf(phase: InputState['phase']): 'plain' | 'claimed' | 'frozen' { /** Whether two projections differ in content (selection and caret excluded). */ function projectionContentChanged(prev: EditorProjection, next: EditorProjection): boolean { if (prev.clipboardText !== next.clipboardText || prev.detectText !== next.detectText) return true + if (prev.command !== next.command) return true if (prev.occurrences.length !== next.occurrences.length) return true return next.occurrences.some((occ, i) => { const old = prev.occurrences[i] @@ -172,7 +173,6 @@ export class SessionInputShell implements SessionInput { onUpdate: () => { this.onEditorUpdate() }, openReference: (source, reference) => this.deps.inputTriggers?.()?.openReference(source, reference) ?? false, - activeClaimToken: () => this.activeClaimToken(), lexicon: () => this.lexicon.getSnapshot(), resolveLexicon: () => this.deps.inputTriggers?.()?.lexicon, }) @@ -183,7 +183,7 @@ export class SessionInputShell implements SessionInput { // ---- editor plumbing ---- - /** Re-project, run the claim watch, publish, and feed trigger tracking after every editor commit. */ + /** Re-project, sync the claim with the leading command chip, publish, and feed trigger tracking after every editor commit. */ private onEditorUpdate(): void { const prev = this.draftEditor.refreshProjection() // Selection-only commits advance neither the revision nor the published @@ -196,7 +196,7 @@ export class SessionInputShell implements SessionInput { this.failedDetached.clear() this.failedRestoreRev = undefined } - this.dispatchRun(({ type: 'draft-changed', draft: this.projection.clipboardText })) + this.dispatchRun(({ type: 'draft-changed', command: this.projection.command })) } const caret = this.projection.caret if (caret !== null) { @@ -449,8 +449,8 @@ export class SessionInputShell implements SessionInput { /** * Apply one command claim (scoped begin-command event listener body): the - * editor replaces [0, span.end) with the claim token, then the machine - * enters claimed. + * editor replaces [0, span.end) with the command chip and its separator, + * then the machine enters claimed. * @param claim - the command claim from the pick path. * @param span - pick-time span snapshot (detect coordinates). * @returns whether the edit applied (phase, span CAS, and leading guard passed). @@ -460,9 +460,10 @@ export class SessionInputShell implements SessionInput { if (phase !== 'plain' && phase !== 'claimed') return false if (span.draftRev !== this.rev) return false // Leading-trigger contract: only whitespace may precede the span; the - // whitespace prefix is dropped so the claimed watch (startsWith) holds. + // whitespace prefix is dropped so the chip takes the leading seat. if (this.projection.detectText.slice(0, span.start).trim() !== '') return false - const applied = this.draftEditor.replaceText({ start: 0, end: span.end }, claim.token) + const tail = this.projection.detectText.slice(span.end, span.end + 1) + const applied = this.draftEditor.insertCommand({ start: 0, end: span.end }, claim, tail) if (!applied) return false this.dispatchRun(({ type: 'claim', claim })) return true @@ -616,20 +617,9 @@ export class SessionInputShell implements SessionInput { // ---- effect executor ---- - /** The claim token the decoration transform styles; null while unclaimed. */ - private activeClaimToken(): string | null { - const core = this.core.state - return (core.phase === 'claimed' || core.phase === 'submitting') && core.claim !== undefined - ? core.claim.token - : null - } - - /** Dispatch + execute, refreshing the claim decoration when the styled token flips. */ + /** Dispatch one machine event and execute its effects. */ private dispatchRun(ev: Parameters[0]): void { - const beforeToken = this.activeClaimToken() - const effects = this.core.dispatch(ev) - this.run(effects) - if (this.activeClaimToken() !== beforeToken) this.draftEditor.refreshClaimDecoration() + this.run(this.core.dispatch(ev)) } private run(effects: readonly InputEffect[]): void { @@ -857,7 +847,7 @@ export class SessionInputShell implements SessionInput { } this.dispatchRun(({ type: 'submit-settled', attempt, ok: outcome.kind === 'success', - draft: this.projection.clipboardText, outcome, + draft: this.projection.clipboardText, command: this.projection.command, outcome, ...(outcome.kind === 'error' && outcome.text === undefined ? { message: 'command failed' } : {}), })) }, @@ -866,7 +856,7 @@ export class SessionInputShell implements SessionInput { const message = error instanceof Error ? error.message : String(error) this.dispatchRun(({ type: 'submit-settled', attempt, ok: false, - draft: this.projection.clipboardText, message, + draft: this.projection.clipboardText, command: this.projection.command, message, })) }, ) diff --git a/packages/client/ui-conversation/src/client/input/machine.ts b/packages/client/ui-conversation/src/client/input/machine.ts index 910a0a66fe..1eec268602 100644 --- a/packages/client/ui-conversation/src/client/input/machine.ts +++ b/packages/client/ui-conversation/src/client/input/machine.ts @@ -27,11 +27,6 @@ function argsAfter(draft: string, token: string): string { return '' } -/** A claimed name may stand alone; arguments require the token's separator. */ -function retainsClaim(draft: string, token: string): boolean { - return draft.startsWith(token) || draft === token.trimEnd() -} - /** The submit-plane slice of the published InputState. */ export interface SubmitSnapshot { readonly phase: InputState['phase'] @@ -75,7 +70,7 @@ export class SubmitMachine { */ dispatch(ev: InputEvent): readonly InputEffect[] { switch (ev.type) { - case 'draft-changed': return this.onDraftChanged(ev.draft) + case 'draft-changed': return this.onDraftChanged(ev.command) case 'claim': return this.onClaim(ev.claim) case 'enter': return this.onEnter(ev.mode, ev.draft, ev.submission) case 'adjudicated': return this.onAdjudicated(ev.attempt, ev.outcome) @@ -88,16 +83,19 @@ export class SubmitMachine { } } - /** The complete command name retains its claim with or without the argument separator. */ - private onDraftChanged(draft: string): readonly InputEffect[] { - if (this.phase === 'claimed' && this.claim !== undefined && !retainsClaim(draft, this.claim.token)) { - this.phase = 'plain' - this.claim = undefined - } + /** + * Outside a transaction the claim follows the document's leading command + * chip: deleting or displacing the chip releases it, and a history restore + * of the chip re-enters claimed. + */ + private onDraftChanged(command: CommandClaim | undefined): readonly InputEffect[] { + if (this.phase !== 'plain' && this.phase !== 'claimed') return [] + this.claim = command + this.phase = command === undefined ? 'plain' : 'claimed' return [] } - /** The editor applied a claim-token replacement; busy phases refuse another claim. */ + /** The editor inserted a command chip; busy phases refuse another claim. */ private onClaim(claim: CommandClaim): readonly InputEffect[] { if (this.phase !== 'plain' && this.phase !== 'claimed') return [] this.claim = claim @@ -204,8 +202,7 @@ export class SubmitMachine { return effects } const text = ev.message ?? ev.outcome?.text - if (ev.draft === flight.attempt.draftSnapshot - && this.claim !== undefined && retainsClaim(ev.draft, this.claim.token)) { + if (ev.draft === flight.attempt.draftSnapshot && this.claim !== undefined && ev.command === this.claim) { this.phase = 'claimed' return text === undefined ? [] : [{ type: 'notice', level: 'error', text }] } diff --git a/packages/client/ui-conversation/src/client/skeleton/InputBar.module.css b/packages/client/ui-conversation/src/client/skeleton/InputBar.module.css index 4676f4cd7d..123a0fcbea 100644 --- a/packages/client/ui-conversation/src/client/skeleton/InputBar.module.css +++ b/packages/client/ui-conversation/src/client/skeleton/InputBar.module.css @@ -202,6 +202,11 @@ content: none; } +/* A selected command chip stands in for the caret until the next key. */ +.input[data-composer-chip-selected] { + caret-color: transparent; +} + .input[data-composer-composing] + .placeholder { visibility: hidden; } diff --git a/packages/client/ui-conversation/tests/editor-focus.client.spec.ts b/packages/client/ui-conversation/tests/editor-focus.client.spec.ts index 56f1902291..7f115ff744 100644 --- a/packages/client/ui-conversation/tests/editor-focus.client.spec.ts +++ b/packages/client/ui-conversation/tests/editor-focus.client.spec.ts @@ -37,7 +37,7 @@ function fixture(start = 1, end = 4) { const onUpdate = vi.fn() const runtime: DraftEditorRuntime = new DraftEditorRuntime({ onUpdate: () => { runtime.refreshProjection(); onUpdate() }, - openReference: () => false, activeClaimToken: () => null, + openReference: () => false, lexicon: () => new Map(), resolveLexicon: () => undefined, }) const dispose = runtime.register() diff --git a/packages/client/ui-conversation/tests/input-bar.client.spec.tsx b/packages/client/ui-conversation/tests/input-bar.client.spec.tsx index 7ba82035d5..f7d55f9375 100644 --- a/packages/client/ui-conversation/tests/input-bar.client.spec.tsx +++ b/packages/client/ui-conversation/tests/input-bar.client.spec.tsx @@ -1425,12 +1425,12 @@ describe('machine pending lock', () => { }) describe('decorations', () => { - /** The claim-token styled leaf (the transform's inline accent color). */ - function tokenSpanOf(container: HTMLElement): HTMLElement | null { - return container.querySelector('[data-lexical-text][style*="business-primary"]') + /** The claimed command chip's host element. */ + function commandChipOf(container: HTMLElement): HTMLElement | null { + return container.querySelector('[data-composer-command]') } - it('claimed token styles the leading leaf and sets the blank-args hint variable', () => { + it('a claim renders the leading command chip and sets the blank-args hint variable', () => { // Dictionary-less stub: an unmatched hint key keeps the machine's raw hint. const { view, shell, textarea } = bench({ t: makeTranslate({}) }) act(() => { @@ -1439,15 +1439,14 @@ describe('decorations', () => { { name: 'goal', token: '/goal ', hint: '目标内容', submit: () => Promise.resolve({ kind: 'success' as const }) }, { start: 0, end: 6, draftRev: shell.snapshot.draftRev }, ) - shell.editor.update(() => {}, { discrete: true }) // flush the queued decoration refresh }) - expect(tokenSpanOf(view.container)?.textContent).toBe('/goal ') + expect(commandChipOf(view.container)?.textContent).toBe('/goal') expect(textarea.style.getPropertyValue('--dsh-composer-hint')).toBe(JSON.stringify('目标内容')) - // Args typed: the hint disappears, the token style stays. - act(() => { shell.setDraft('/goal 发布') }) - act(() => { shell.editor.update(() => {}, { discrete: true }) }) // flush the queued decoration refresh + // Args typed: the hint disappears, the chip stays. + act(() => { shell.actions.insertText('发布', shell.actions.captureInsertion()) }) + expect(shell.snapshot.draft).toBe('/goal 发布') expect(textarea.style.getPropertyValue('--dsh-composer-hint')).toBe('') - expect(tokenSpanOf(view.container)).not.toBeNull() + expect(commandChipOf(view.container)).not.toBeNull() }) it('a locale entry for the claimed command overrides the raw claim hint (trailing-space token)', () => { diff --git a/packages/client/ui-conversation/tests/input-matrix.client.spec.tsx b/packages/client/ui-conversation/tests/input-matrix.client.spec.tsx index e7c997cb61..81bbe030b3 100644 --- a/packages/client/ui-conversation/tests/input-matrix.client.spec.tsx +++ b/packages/client/ui-conversation/tests/input-matrix.client.spec.tsx @@ -1,7 +1,7 @@ // @vitest-environment jsdom /** * Impact-matrix projection tests (row by row): what each - * phase projects onto the InputBar — enter routing, visuals (token color / + * phase projects onto the InputBar — enter routing, visuals (command chip / * hint / pending), edit freedom, and the published currency's claim seat. * React over jsdom per the client testing discipline; the machine is real. */ @@ -121,7 +121,11 @@ function bench(over?: { ) }) } - return { view, textarea, shell, wiring, sink, claim, serialize, release } + /** Type at the live caret, as keyboard input after the claimed chip does. */ + const typeAtCaret = (text: string) => { + act(() => { shell.actions.insertText(text, shell.actions.captureInsertion()) }) + } + return { view, textarea, shell, wiring, sink, claim, typeAtCaret, serialize, release } } describe('matrix row: plain', () => { @@ -158,26 +162,29 @@ describe('matrix row: plain', () => { }) describe('matrix row: claimed', () => { - it('publishes the claim currency, colors the token, hints while args are blank, and edits stay free', () => { - const { view, textarea, shell, claim } = bench() + it('publishes the claim currency, renders the command chip, hints while args are blank, and edits stay free', () => { + const { view, textarea, shell, claim, typeAtCaret } = bench() claim() - act(() => { shell.editor.update(() => {}, { discrete: true }) }) // flush the queued decoration refresh expect(shell.snapshot.claim).toEqual({ name: 'goal', token: '/goal ', hint: '目标' }) - expect(view.container.querySelector('[data-lexical-text][style*="business-primary"]')?.textContent).toBe('/goal ') + const chip = view.container.querySelector('[data-composer-command="goal"]') + expect(chip?.textContent).toBe('/goal') + expect(chip?.getAttribute('contenteditable')).toBe('false') + expect(shell.snapshot.draft).toBe('/goal ') // The zh dictionary owns a hint.goal entry, which overrides the raw claim hint (production behavior). expect(textarea.style.getPropertyValue('--dsh-composer-hint')).toBe(JSON.stringify('输入目标，智能体将持续执行')) expect(textarea.getAttribute('contenteditable')).toBe('true') - // Free editing beyond the token: hint drops, claim holds. - act(() => { shell.setDraft('/goal 发布版本') }) + // Free editing after the chip: hint drops, claim holds. + typeAtCaret('发布版本') + expect(shell.snapshot.draft).toBe('/goal 发布版本') expect(shell.snapshot.phase).toBe('claimed') expect(textarea.style.getPropertyValue('--dsh-composer-hint')).toBe('') }) it('enter routes to claim.submit (command lane, never the queue sink)', async () => { const submit = vi.fn(() => Promise.resolve({ kind: 'success' as const, text: '完成', source: 'command', name: 'goal' })) - const { view, textarea, shell, sink, claim } = bench({ submit }) + const { view, textarea, shell, sink, claim, typeAtCaret } = bench({ submit }) claim() - act(() => { shell.setDraft('/goal 发布') }) + typeAtCaret('发布') fireEvent.keyDown(textarea, { key: 'Enter' }) expect(sink).not.toHaveBeenCalled() await vi.waitFor(() => { expect(submit).toHaveBeenCalledWith('发布', SCTX, []) }) @@ -186,14 +193,13 @@ describe('matrix row: claimed', () => { expect(view.getByText('完成')).toBeTruthy() }) - it('backspacing the token auto-releases to plain and the visuals vanish (scenario H)', () => { + it('replacing the whole draft removes the chip and releases to plain', () => { const { view, shell, claim } = bench() claim() - act(() => { shell.setDraft('/goa 发布') }) // token broken + act(() => { shell.setDraft('/goal 发布') }) expect(shell.snapshot.phase).toBe('plain') expect(shell.snapshot.claim).toBeUndefined() - act(() => { shell.editor.update(() => {}, { discrete: true }) }) // flush the queued decoration refresh - expect(view.container.querySelector('[data-lexical-text][style*="business-primary"]')).toBeNull() + expect(view.container.querySelector('[data-composer-command]')).toBeNull() }) }) diff --git a/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx b/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx index ca03589a52..d30124bf24 100644 --- a/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx +++ b/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx @@ -12,7 +12,7 @@ import './control-row-dom.ts' import type { GlobalStandardProps } from '@deepseek-ai/dsh-client-ui-slots' import { afterEach, describe, expect, it, onTestFinished, vi } from 'vitest' import { act, cleanup, fireEvent, render } from '@testing-library/react' -import { $getRoot, $isElementNode, $isTextNode } from 'lexical' +import { $createTextNode, $getRoot, $isElementNode, DELETE_CHARACTER_COMMAND, UNDO_COMMAND } from 'lexical' import type { SessionSnapshot } from '@deepseek-ai/dsh-api-session-controller/client' import { createSnapshotStore } from '@deepseek-ai/dsh-client-store' import { InputTriggerService } from '@deepseek-ai/dsh-client-ui-input-trigger/client' @@ -205,7 +205,32 @@ async function scopedBench(register?: (inputTriggers: InputTriggerService) => vo const type = (text: string): void => { act(() => { shell.setDraft(text) }) } - return { runtime, inputTriggers, controller, shell, wiring, view, textarea, type, sink, serialize, release } + /** Press Space and let the claim the key handler applied commit (Lexical commits it on the next microtask). */ + const space = async (): Promise => { + fireEvent.keyDown(textarea, { key: ' ', keyCode: 32 }) + await act(async () => {}) + } + /** Type at the live caret, as keyboard input after the claimed chip does. */ + const append = (text: string): void => { + act(() => { shell.actions.insertText(text, shell.actions.captureInsertion()) }) + } + /** Dispatch one character deletion and commit it (the command's update lands on the next flush). */ + const deleteCharacter = (isBackward: boolean): void => { + act(() => { + shell.editor.dispatchCommand(DELETE_CHARACTER_COMMAND, isBackward) + shell.editor.update(() => {}, { discrete: true }) + }) + } + /** Delete one detect span through the editor, as a Backspace over it does. */ + const erase = (start: number, end: number): void => { + act(() => { + shell.editor.update(() => { $replaceDetectSpanWithText({ start, end }, '') }, { discrete: true }) + }) + } + return { + runtime, inputTriggers, controller, shell, wiring, view, textarea, + type, space, append, erase, deleteCharacter, sink, serialize, release, + } } async function bench(executeImpl?: (line: string) => Promise) { @@ -230,12 +255,11 @@ describe('scenario A: menu-pick /goal, type args, enter submits', () => { act(() => { b.controller.pick('command', 0) }) expect(b.shell.snapshot.phase).toBe('claimed') expect(b.shell.snapshot.draft).toBe('/goal ') - act(() => { b.shell.editor.update(() => {}, { discrete: true }) }) // flush the queued decoration refresh - expect(b.view.container.querySelector('[data-lexical-text][style*="business-primary"]')?.textContent).toBe('/goal ') + expect(b.view.container.querySelector('[data-composer-command]')?.textContent).toBe('/goal') // The zh dictionary owns a hint.goal entry, which overrides the machine's raw hint (production behavior). expect(b.textarea.style.getPropertyValue('--dsh-composer-hint')).toBe(JSON.stringify('输入目标，智能体将持续执行')) // Continue typing args; hint drops; claim holds. - b.type('/goal 发布 v1') + b.append('发布 v1') expect(b.shell.snapshot.phase).toBe('claimed') // Enter: submitting → command execute → commit clears. fireEvent.keyDown(b.textarea, { key: 'Enter' }) @@ -316,71 +340,170 @@ describe('scenario: images ride an accepting command through the real pipeline', }) }) -describe('scenario H: backspace breaks the token', () => { - it.each(['goal', '目标', 'plan', '计划', 'feedback', '反馈'])('keeps /%s claimed when its arguments and separator are deleted', async (name) => { +describe('scenario H: the command chip deletes whole', () => { + it.each(['goal', '目标', 'plan', '计划', 'feedback', '反馈'])('keeps /%s claimed until the chip itself is deleted', async (name) => { const { source } = commandSource([{ name, description: name, input: { hint: '目标内容' } }], () => Promise.resolve({ kind: 'success' })) const b = await scopedBench((triggers) => { triggers.registerSource(source) }) b.type(`/${name}`) - fireEvent.keyDown(b.textarea, { key: ' ', keyCode: 32 }) + await b.space() expect(b.shell.snapshot.phase).toBe('claimed') - b.type(`/${name} 这是目标`) + b.append('这是目标') for (let i = 0; i < 5; i++) { - act(() => { - b.shell.editor.update(() => { - const end = b.shell.snapshot.draft.length - $replaceDetectSpanWithText({ start: end - 1, end }, '') - }, { discrete: true }) - }) + const { end } = b.shell.caretSpan() + b.erase(end - 1, end) } expect(b.shell.snapshot.draft).toBe(`/${name}`) expect(b.shell.snapshot.phase).toBe('claimed') - expect(b.view.container.querySelector('[data-lexical-text][style*="business-primary"]')?.textContent).toBe(`/${name}`) - b.type(`/${name} `) + expect(b.view.container.querySelector('[data-composer-command]')?.textContent).toBe(`/${name}`) + b.append(' ') await act(async () => {}) expect(b.shell.snapshot.phase).toBe('claimed') expect(b.shell.snapshot.draft).toBe(`/${name} `) - expect(b.view.container.querySelector('[data-lexical-text][style*="business-primary"]')?.textContent).toBe(`/${name} `) - b.type(`/${name}x`) + b.erase(0, 1) expect(b.shell.snapshot.phase).toBe('plain') + expect(b.shell.snapshot.draft).toBe(' ') + expect(b.view.container.querySelector('[data-composer-command]')).toBeNull() }) - it('keeps text typed at the end of the claimed token when the browser inserts it natively', async () => { + it('the first deletion at the chip edge selects the chip and the second removes it', async () => { const b = await bench() b.type('/goal') await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) - fireEvent.keyDown(b.textarea, { key: ' ', keyCode: 32 }) + await b.space() + b.append('发布') + for (let i = 0; i < 3; i++) { + const { end } = b.shell.caretSpan() + b.erase(end - 1, end) + } + expect(b.shell.snapshot.draft).toBe('/goal') + const chip = (): Element | null => b.view.container.querySelector('[data-composer-command]') + b.deleteCharacter(true) + expect(chip()?.hasAttribute('data-selected')).toBe(true) + expect(b.textarea.hasAttribute('data-composer-chip-selected')).toBe(true) expect(b.shell.snapshot.phase).toBe('claimed') - b.type('/goal 发布') - // A caret at the token/argument boundary sits at the end of the styled - // token node; native insertion writes there and Lexical syncs the DOM - // text with setTextContent plus a caret select. + b.deleteCharacter(true) + expect(chip()).toBeNull() + expect(b.shell.snapshot.phase).toBe('plain') + expect(b.textarea.hasAttribute('data-composer-chip-selected')).toBe(false) + }) + + it('another key or typing clears a selected chip without deleting it', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.erase(1, 2) + const chip = (): Element | null => b.view.container.querySelector('[data-composer-command]') + b.deleteCharacter(true) + expect(chip()?.hasAttribute('data-selected')).toBe(true) + fireEvent.keyDown(b.textarea, { key: 'Shift' }) + expect(chip()?.hasAttribute('data-selected')).toBe(true) + fireEvent.keyDown(b.textarea, { key: 'Escape' }) + expect(chip()?.hasAttribute('data-selected')).toBe(false) + expect(b.textarea.hasAttribute('data-composer-chip-selected')).toBe(false) + b.deleteCharacter(true) + expect(chip()?.hasAttribute('data-selected')).toBe(true) + b.append('发布') + expect(chip()?.hasAttribute('data-selected')).toBe(false) + expect(b.shell.snapshot.draft).toBe('/goal发布') + expect(b.shell.snapshot.phase).toBe('claimed') + }) + + it('Delete before the chip selects it first and removes it on the second press', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.append('发布') act(() => { b.shell.editor.update(() => { const paragraph = $getRoot().getFirstChild() - const token = $isElementNode(paragraph) ? paragraph.getFirstChild() : null - if (!$isTextNode(token)) throw new Error('missing token node') - token.setTextContent('/goal 立即') - token.select(8, 8) + if (!$isElementNode(paragraph)) throw new Error('missing paragraph') + paragraph.select(0, 0) }, { discrete: true }) }) - expect(b.shell.snapshot.draft).toBe('/goal 立即发布') - expect(b.shell.snapshot.phase).toBe('claimed') - act(() => { b.shell.editor.update(() => {}, { discrete: true }) }) - expect(b.view.container.querySelector('[data-lexical-text][style*="business-primary"]')?.textContent).toBe('/goal ') + const chip = (): Element | null => b.view.container.querySelector('[data-composer-command]') + b.deleteCharacter(false) + expect(chip()?.hasAttribute('data-selected')).toBe(true) + b.deleteCharacter(false) + expect(chip()).toBeNull() + expect(b.shell.snapshot.draft).toBe(' 发布') + expect(b.shell.snapshot.phase).toBe('plain') }) - it('claim releases automatically; the enter after that goes through adjudication again', async () => { + it('keeps text written at the chip boundary out of the chip', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.append('发布') + // The caret right after the chip: Lexical writes into the argument text. + act(() => { + b.shell.editor.update(() => { + const paragraph = $getRoot().getFirstChild() + if (!$isElementNode(paragraph)) throw new Error('missing paragraph') + paragraph.select(1, 1).insertText('立即') + }, { discrete: true }) + }) + expect(b.shell.snapshot.draft).toBe('/goal立即 发布') + expect(b.shell.snapshot.phase).toBe('claimed') + expect(b.view.container.querySelector('[data-composer-command]')?.textContent).toBe('/goal') + }) + + it('deleting the chip releases the claim; the arguments left behind send as a plain message', async () => { const b = await bench() b.type('/goal') await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) // Space adjudication claims (space column, leadingInput). - fireEvent.keyDown(b.textarea, { key: ' ', keyCode: 32 }) + await b.space() expect(b.shell.snapshot.phase).toBe('claimed') - // Backspace into the token: watch break → plain, visuals gone. - b.type('/goa ') + b.append('发布') + b.erase(0, 1) expect(b.shell.snapshot.phase).toBe('plain') - expect(b.view.container.querySelector('[data-lexical-text][style*="business-primary"]')).toBeNull() + expect(b.shell.snapshot.draft).toBe(' 发布') + expect(b.view.container.querySelector('[data-composer-command]')).toBeNull() + fireEvent.keyDown(b.textarea, { key: 'Enter' }) + expect(b.sink).toHaveBeenCalledWith('发布', [], 'queue', expect.any(AbortSignal)) + expect(b.execute).not.toHaveBeenCalled() + }) + + it('typing before the chip turns it into its literal token and releases the claim', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.append('发布') + act(() => { + b.shell.editor.update(() => { + const paragraph = $getRoot().getFirstChild() + if (!$isElementNode(paragraph)) throw new Error('missing paragraph') + paragraph.getFirstChild()?.insertBefore($createTextNode('看看 ')) + }, { discrete: true }) + }) + expect(b.shell.snapshot.phase).toBe('plain') + expect(b.shell.snapshot.draft).toBe('看看 /goal 发布') + expect(b.view.container.querySelector('[data-composer-command]')).toBeNull() + }) + + it('undoing the chip deletion restores the chip and its claim', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.append('发布') + b.erase(0, 1) + expect(b.shell.snapshot.phase).toBe('plain') + act(() => { + b.shell.editor.dispatchCommand(UNDO_COMMAND, undefined) + b.shell.editor.update(() => {}, { discrete: true }) // the history restore commits on the next flush + }) + expect(b.shell.snapshot.phase).toBe('claimed') + expect(b.shell.snapshot.claim?.name).toBe('goal') + expect(b.shell.snapshot.draft).toBe('/goal 发布') + expect(b.view.container.querySelector('[data-composer-command]')?.textContent).toBe('/goal') + fireEvent.keyDown(b.textarea, { key: 'Enter' }) + await vi.waitFor(() => { expect(b.execute).toHaveBeenCalledWith('/goal 发布', []) }) }) }) diff --git a/packages/client/ui-conversation/tests/lexical-editor-core.client.spec.tsx b/packages/client/ui-conversation/tests/lexical-editor-core.client.spec.tsx index cdee0965ba..0922129b23 100644 --- a/packages/client/ui-conversation/tests/lexical-editor-core.client.spec.tsx +++ b/packages/client/ui-conversation/tests/lexical-editor-core.client.spec.tsx @@ -1,7 +1,7 @@ // @vitest-environment jsdom /** - * Composer editor pure core: ReferenceChipNode semantics, the three-view - * projections, and detect-span application. Headless editors drive every + * Composer editor pure core: ReferenceChipNode and CommandChipNode + * semantics, the three-view projections, and detect-span application. Headless editors drive every * case — no React tree, no contenteditable; the chip's visual face has its * own component spec. */ @@ -13,11 +13,14 @@ import { $isTextNode, $setSelection, } from 'lexical' import type { ReferenceInsert } from '../src/client/contract/draft-editor.ts' +import type { CommandClaim } from '../src/client/contract/input.ts' import { $createReferenceChipNode, $isReferenceChipNode, ReferenceChipNode, } from '../src/client/input/editor/chip-node.tsx' -import { registerClaimDecoration } from '../src/client/input/editor/claim-decor.ts' -import { registerTextRefDecoration, TextRefNode } from '../src/client/input/editor/text-ref.ts' +import { + $createCommandChipNode, $isCommandChipNode, CommandChipNode, registerCommandChipSeat, +} from '../src/client/input/editor/command-chip-node.tsx' +import { TextRefNode } from '../src/client/input/editor/text-ref.ts' import type { SerializedReferenceChipNode } from '../src/client/input/editor/chip-node.tsx' import { $composerLayout, $detectOffsetOfPoint, $projectComposer, ATOMIC_CHAR, @@ -44,7 +47,7 @@ const SKILL_REF: ReferenceInsert = { function makeEditor(): LexicalEditor { return createHeadlessEditor({ namespace: 'core-spec', - nodes: [ReferenceChipNode, TextRefNode], + nodes: [ReferenceChipNode, CommandChipNode, TextRefNode], onError: (error) => { throw error }, }) } @@ -425,44 +428,109 @@ describe('detect-span application', () => { }) }) -describe('claim precedence over text-ref entities', () => { - const TOKEN_STYLE = 'color: var(--dsw-alias-state-business-primary)' - const LEXICON: ReadonlyMap<'/' | '@', readonly string[]> = new Map([['/', ['plan']]]) +const GOAL_CLAIM: CommandClaim = { + name: 'goal', + token: '/goal ', + submit: () => Promise.resolve({ kind: 'success' }), +} - it.each(['/plan', '/plan '])('keeps the lexicon-listed %j claim plain and accent-styled until release', (token) => { +/** Seed one paragraph opening with the command chip, followed by `rest`. */ +function seedCommand(editor: LexicalEditor, rest = ' ship'): void { + editor.update(() => { + const paragraph = $createParagraphNode() + paragraph.append($createCommandChipNode(GOAL_CLAIM)) + if (rest !== '') paragraph.append($createTextNode(rest)) + $getRoot().clear().append(paragraph) + }, { discrete: true }) +} + +/** The first paragraph of the document. */ +function $firstParagraph(): ParagraphNode { + return $getRoot().getFirstChild() as ParagraphNode +} + +describe('CommandChipNode', () => { + it('projects as one atomic character, the token without its separator, and the leading claim', () => { const editor = makeEditor() - let claim: string | null = token - registerClaimDecoration(editor, () => claim) - registerTextRefDecoration(editor, () => LEXICON, () => claim) + seedCommand(editor) + editor.read(() => { + const projection = $projectComposer(idAssigner()) + expect(projection.detectText).toBe(`${ATOMIC_CHAR} ship`) + expect(projection.clipboardText).toBe('/goal ship') + expect(projection.occurrences).toEqual([]) + expect(projection.command).toBe(GOAL_CLAIM) + }) + }) + + it('keeps its claim across writable clones', () => { + const editor = makeEditor() + seedCommand(editor) editor.update(() => { - const paragraph = $createParagraphNode() - paragraph.append($createTextNode('/plan rest')) - $getRoot().clear().append(paragraph) + const chip = $firstParagraph().getFirstChild() + if (!$isCommandChipNode(chip)) throw new Error('missing chip') + chip.markDirty() }, { discrete: true }) - const leaf = (): { type: string; style: string; text: string } | null => - editor.getEditorState().read(() => { - const first = ($getRoot().getFirstChild() as ParagraphNode).getFirstChild() - return first === null ? null : { - type: first.getType(), - style: $isTextNode(first) ? first.getStyle() : '', - text: first.getTextContent(), - } - }) - // Claimed: the entity transform yields the seat, the claim transform styles it. - expect(leaf()).toEqual({ type: 'text', style: TOKEN_STYLE, text: token }) - // Released (the shell's refresh nudges the seat dirty): the entity captures it. - claim = null + editor.read(() => { + const chip = $firstParagraph().getFirstChild() + expect($isCommandChipNode(chip) && chip.getClaim()).toBe(GOAL_CLAIM) + }) + }) + + it('serializes the token text and imports it back as plain text', () => { + const editor = makeEditor() + seedCommand(editor, '') + const json = editor.getEditorState().toJSON() + expect(json.root.children[0]).toMatchObject({ children: [{ type: 'command-chip', text: '/goal' }] }) + editor.parseEditorState(json).read(() => { + const first = $firstParagraph().getFirstChild() + expect($isTextNode(first) && first.getTextContent()).toBe('/goal') + }) + }) + + it('becomes its literal token when text lands before it', () => { + const editor = makeEditor() + registerCommandChipSeat(editor) + seedCommand(editor) editor.update(() => { - const first = ($getRoot().getFirstChild() as ParagraphNode).getFirstChild() - if ($isTextNode(first)) first.markDirty() + $firstParagraph().getFirstChild()?.insertBefore($createTextNode('x')) }, { discrete: true }) - expect(leaf()).toEqual({ type: 'composer-text-ref', style: '', text: '/plan' }) - // Re-claimed: the entity reverts to plain text and the warn style returns. - claim = token + editor.read(() => { + const projection = $projectComposer(idAssigner()) + expect(projection.detectText).toBe('x/goal ship') + expect(projection.command).toBeUndefined() + }) + }) + + it('becomes its literal token behind a line break or on a later paragraph', () => { + const editor = makeEditor() + registerCommandChipSeat(editor) + seedCommand(editor) editor.update(() => { - const first = ($getRoot().getFirstChild() as ParagraphNode).getFirstChild() - if ($isTextNode(first)) first.markDirty() + $firstParagraph().getFirstChild()?.insertBefore($createLineBreakNode()) }, { discrete: true }) - expect(leaf()).toEqual({ type: 'text', style: TOKEN_STYLE, text: token }) + editor.read(() => { + expect($projectComposer(idAssigner()).detectText).toBe('\n/goal ship') + }) + seedCommand(editor) + editor.update(() => { + $firstParagraph().insertBefore($createParagraphNode()) + }, { discrete: true }) + editor.read(() => { + const projection = $projectComposer(idAssigner()) + expect(projection.detectText).toBe('\n/goal ship') + expect(projection.command).toBeUndefined() + }) + }) + + it('stays a chip on the leading seat', () => { + const editor = makeEditor() + registerCommandChipSeat(editor) + seedCommand(editor) + editor.update(() => { + $firstParagraph().getLastChild()?.remove() + }, { discrete: true }) + editor.read(() => { + expect($projectComposer(idAssigner()).command).toBe(GOAL_CLAIM) + }) }) }) diff --git a/packages/client/ui-conversation/tests/submit-machine.client.spec.ts b/packages/client/ui-conversation/tests/submit-machine.client.spec.ts index 9bda0536aa..4ecb8c2e92 100644 --- a/packages/client/ui-conversation/tests/submit-machine.client.spec.ts +++ b/packages/client/ui-conversation/tests/submit-machine.client.spec.ts @@ -1,9 +1,9 @@ /** * SubmitMachine behavior: enter routing, adjudication outcomes, the claimed - * lifecycle and its integrity watch, settlement (commit-draft and claim - * re-entry decisions), anti-backwash, and per-session isolation. Text-edit - * semantics live in the editor (lexical-editor-core spec) — the machine only - * observes drafts through event payloads. + * lifecycle that follows the leading command chip, settlement (commit-draft + * and claim re-entry decisions), anti-backwash, and per-session isolation. + * Text-edit semantics live in the editor (lexical-editor-core spec) — the + * machine only observes the document through event payloads. */ import { describe, expect, it } from 'vitest' import type { CommandClaim } from '../src/client/contract/input.ts' @@ -192,41 +192,45 @@ describe('submit-machine: claimed lifecycle', () => { expect(m.state.phase).toBe('submitting') }) - it('editing the command name releases back to plain', () => { + it('holds the claim while the leading command chip stays and releases it when the chip goes', () => { const m = new SubmitMachine() - m.dispatch({ type: 'claim', claim: claimOf('goal') }) - m.dispatch({ type: 'draft-changed', draft: '/goal args fine' }) + const claim = claimOf('goal') + m.dispatch({ type: 'claim', claim }) + m.dispatch({ type: 'draft-changed', command: claim }) expect(m.state.phase).toBe('claimed') - m.dispatch({ type: 'draft-changed', draft: '/goa' }) + m.dispatch({ type: 'draft-changed', command: undefined }) expect(m.state.phase).toBe('plain') expect(m.state.claim).toBeUndefined() }) + it('re-enters claimed when a history restore brings the command chip back', () => { + const m = new SubmitMachine() + const claim = { ...claimOf('goal'), token: '/目标 ' } + m.dispatch({ type: 'draft-changed', command: claim }) + expect(m.state.phase).toBe('claimed') + expect(m.state.claim?.token).toBe('/目标 ') + }) + it.each([ ['goal', '/goal '], ['goal', '/目标 '], ['plan', '/plan '], ['plan', '/计划 '], ['feedback', '/feedback '], ['feedback', '/反馈 '], - ])('retains %s as %s without its separator and submits an empty argument', (name, token) => { + ])('submits %s claimed as %s without its separator with an empty argument', (name, token) => { const m = new SubmitMachine() - m.dispatch({ type: 'claim', claim: { ...claimOf(name), token } }) - for (const draft of [token + '这是目标', token, token.trimEnd(), token, token.trimEnd()]) { - m.dispatch({ type: 'draft-changed', draft }) - expect(m.state.phase).toBe('claimed') - expect(m.state.claim?.name).toBe(name) - } + const claim = { ...claimOf(name), token } + m.dispatch({ type: 'claim', claim }) const fx = m.dispatch({ type: 'enter', mode: 'queue', draft: token.trimEnd() }) const begin = effectAt(fx, 0, 'begin-submit') expect(begin.args).toBe('') - m.dispatch({ type: 'submit-settled', attempt: begin.attempt, ok: false, draft: token.trimEnd() }) + m.dispatch({ type: 'submit-settled', attempt: begin.attempt, ok: false, draft: token.trimEnd(), command: claim }) expect(m.state.phase).toBe('claimed') expect(m.state.claim?.name).toBe(name) }) - it.each(['/目', '/目标x', '/目标/文件', '', '看看 /目标'])('releases a goal claim for %j', (draft) => { + it('reads arguments typed directly against the chip without a separator', () => { const m = new SubmitMachine() m.dispatch({ type: 'claim', claim: { ...claimOf('goal'), token: '/目标 ' } }) - m.dispatch({ type: 'draft-changed', draft }) - expect(m.state.phase).toBe('plain') - expect(m.state.claim).toBeUndefined() + const fx = m.dispatch({ type: 'enter', mode: 'queue', draft: '/目标发布' }) + expect(effectAt(fx, 0, 'begin-submit').args).toBe('发布') }) it('explicit release returns to plain when nothing is in flight', () => { @@ -250,7 +254,7 @@ describe('submit-machine: submitting transaction', () => { const m = new SubmitMachine() enterSubmitting(m, 'goal', 'x') expect(m.dispatch({ type: 'enter', mode: 'queue', draft: '/goal x' })).toEqual([]) - m.dispatch({ type: 'draft-changed', draft: 'typed during flight' }) + m.dispatch({ type: 'draft-changed', command: undefined }) expect(m.state.phase).toBe('submitting') }) @@ -258,7 +262,7 @@ describe('submit-machine: submitting transaction', () => { const m = new SubmitMachine() const { attempt } = enterSubmitting(m, 'goal', 'x') const fx = m.dispatch({ - type: 'submit-settled', attempt, ok: true, draft: '/goal x', + type: 'submit-settled', attempt, ok: true, draft: '/goal x', command: undefined, outcome: { kind: 'success', text: 'goal saved' }, }) expect(effectAt(fx, 0, 'commit-draft').retainSuffixOf).toBe('/goal x') @@ -271,49 +275,48 @@ describe('submit-machine: submitting transaction', () => { const m = new SubmitMachine() const { attempt } = enterSubmitting(m, 'goal', 'x') const fx = m.dispatch({ - type: 'submit-settled', attempt, ok: false, draft: 'deviated', + type: 'submit-settled', attempt, ok: false, draft: 'deviated', command: undefined, outcome: { kind: 'error', text: 'rejected' }, }) expect(effectAt(fx, 0, 'notice')).toMatchObject({ level: 'error', text: 'rejected' }) expect(m.state.phase).toBe('plain') }) - it('rollback with an undeviated draft keeps the claim and re-enters claimed', () => { + it('rollback with an undeviated draft and chip keeps the claim and re-enters claimed', () => { const m = new SubmitMachine() - const { attempt } = enterSubmitting(m, 'goal', 'x') - m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: '/goal x', message: 'transport' }) + const { attempt, claim } = enterSubmitting(m, 'goal', 'x') + m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: '/goal x', command: claim, message: 'transport' }) expect(m.state.phase).toBe('claimed') expect(m.state.claim?.token).toBe('/goal ') }) it('rollback with a deviated draft only notices — the newer input wins', () => { const m = new SubmitMachine() - const { attempt } = enterSubmitting(m, 'goal', 'x') - const fx = m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: 'rewritten', message: 'transport' }) + const { attempt, claim } = enterSubmitting(m, 'goal', 'x') + const fx = m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: 'rewritten', command: claim, message: 'transport' }) expect(effectAt(fx, 0, 'notice')).toMatchObject({ level: 'error', text: 'transport' }) expect(m.state.phase).toBe('plain') expect(m.state.claim).toBeUndefined() }) - it('enter-path rollback cannot re-enter claimed when the snapshot never carried the bare token prefix', () => { + it('enter-path rollback returns to plain because an adjudicated claim has no command chip', () => { const m = new SubmitMachine() - const attempt = enterAdjudicating(m, ' /goal x') + const attempt = enterAdjudicating(m, '/goal x') m.dispatch({ type: 'adjudicated', attempt, outcome: { claim: claimOf('goal') } }) - m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: ' /goal x', message: 'nope' }) - // The snapshot carries leading whitespace the token never had: plain, claim cleared. + m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: '/goal x', command: undefined, message: 'nope' }) expect(m.state.phase).toBe('plain') expect(m.state.claim).toBeUndefined() }) it('a stale settle after rollback + resubmit is dropped (anti-backwash)', () => { const m = new SubmitMachine() - const { attempt: first } = enterSubmitting(m, 'goal', 'x') - m.dispatch({ type: 'submit-settled', attempt: first, ok: false, draft: '/goal x', message: 'try again' }) + const { attempt: first, claim } = enterSubmitting(m, 'goal', 'x') + m.dispatch({ type: 'submit-settled', attempt: first, ok: false, draft: '/goal x', command: claim, message: 'try again' }) const fx = m.dispatch({ type: 'enter', mode: 'queue', draft: '/goal x' }) const second = effectAt(fx, 0, 'begin-submit').attempt - expect(m.dispatch({ type: 'submit-settled', attempt: first, ok: true, draft: '/goal x' })).toEqual([]) + expect(m.dispatch({ type: 'submit-settled', attempt: first, ok: true, draft: '/goal x', command: claim })).toEqual([]) expect(m.state.phase).toBe('submitting') - m.dispatch({ type: 'submit-settled', attempt: second, ok: true, draft: '/goal x' }) + m.dispatch({ type: 'submit-settled', attempt: second, ok: true, draft: '/goal x', command: claim }) expect(m.state.phase).toBe('plain') }) @@ -322,7 +325,7 @@ describe('submit-machine: submitting transaction', () => { const { attempt } = enterSubmitting(m, 'goal', 'x') m.dispatch({ type: 'release' }) expect(attempt.signal.aborted).toBe(true) - expect(m.dispatch({ type: 'submit-settled', attempt, ok: true, draft: '' })).toEqual([]) + expect(m.dispatch({ type: 'submit-settled', attempt, ok: true, draft: '', command: undefined })).toEqual([]) expect(m.state.phase).toBe('plain') }) @@ -343,7 +346,7 @@ describe('submit-machine: per-session isolation', () => { const { attempt } = enterSubmitting(a, 'goal', 'x') const fx = b.dispatch({ type: 'enter', mode: 'queue', draft: 'hello' }) expect(effectAt(fx, 0, 'default-sink').draft).toBe('hello') - a.dispatch({ type: 'submit-settled', attempt, ok: true, draft: '/goal x' }) + a.dispatch({ type: 'submit-settled', attempt, ok: true, draft: '/goal x', command: undefined }) expect(a.state.phase).toBe('plain') expect(b.state.phase).toBe('plain') }) From 510cd4cdcce5ba294429990a725bc3b3934e5938 Mon Sep 17 00:00:00 2001 From: Yichen Jiang Date: Fri, 9 Oct 2026 14:52:21 +0800 Subject: [PATCH 007/162] fix(web): cover every deletion path in two-step chip deletion Decide the command chip's selecting step on deletion input events as well as the character-deletion command, so iOS ranged deleteContentBackward no longer removes the chip in one press. A selected chip now clears only after a keydown that produced no character deletion, which keeps platform shortcuts such as macOS Ctrl+H working. A failed command submission settles to the leading chip like every other transition, so submit-settled no longer carries the draft. The claim event reuses the same handler. The seat rule runs as chip and paragraph transforms instead of walking the whole document on every update. Document the reload limitation under Known Limitations and assert that a chip ending its block does not grow the composer. --- ...2026-10-09-composer-command-chip.i18n.yaml | 4 +- .../2026-10-09-composer-command-chip.md | 6 +- .../2026-10-09-composer-command-chip.zh.md | 6 +- apps/web/tests/lifecycle-chrome.e2e.ts | 4 + .../client/ui-conversation/README.i18n.yaml | 4 +- packages/client/ui-conversation/README.md | 1 + packages/client/ui-conversation/README.zh.md | 1 + .../src/client/contract/input.ts | 3 +- .../client/input/editor/command-chip-node.tsx | 38 ++++++---- .../input/editor/command-chip-selection.ts | 74 +++++++++++++++---- .../src/client/input/facade.ts | 4 +- .../src/client/input/machine.ts | 28 +++---- .../tests/input-scenarios.client.spec.tsx | 61 ++++++++++++++- .../tests/submit-machine.client.spec.ts | 29 ++++---- 14 files changed, 183 insertions(+), 80 deletions(-) diff --git a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml index 00e4d39595..03f9288c03 100644 --- a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml +++ b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.i18n.yaml @@ -9,8 +9,8 @@ en: 831c97e0de131639 zh: ea35b4b179ff33f0 /agent-note-claimed-commands-as-atomic-composer-chips/decision: - en: bde405a14ce2011e - zh: 272df3039ae260df + en: 475caae20c483a3e + zh: c8563aa73ed777ab /agent-note-claimed-commands-as-atomic-composer-chips/alternatives-considered: en: 9c694a21237d7772 zh: b35ba7bf8b73b679 diff --git a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md index 366531ca5a..b0a1257339 100644 --- a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md +++ b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.md @@ -14,9 +14,9 @@ A claim replaces the leading trigger span with `CommandChipNode`, an atomic inli - The node carries the `CommandClaim`, including its submit closure. History restores nodes with their fields, so undoing a chip deletion restores the claim. JSON export writes the token text; JSON import yields a plain text node because the closure has no JSON form. - The detect projection counts the chip as one U+FFFC, so trigger detection never re-reads the command name. The clipboard projection is the token without its separator. The chip is not a reference occurrence. `EditorProjection.command` is the claim of a chip that opens the document. -- Outside a submit transaction, the machine's claim follows the document: `draft-changed` carries the leading chip's claim, so deleting or displacing the chip releases the claim and a history restore re-enters claimed. A failed command submission returns to claimed only when the draft is unchanged and the same chip still leads. A claim adjudicated from typed text at Enter has no chip and returns to plain on failure. -- A root transform keeps the chip on the leading seat. A chip with a preceding sibling or inside a later block becomes its literal token, which releases the claim as typing before the old token did. -- Deletion takes two steps. A character deletion that reaches the chip from a collapsed caret marks it selected (`data-selected` on its host; the root's `data-composer-chip-selected` hides the caret), and the next deletion in the same direction removes it whole. Any other key, a click, blur, or a caret that leaves the chip clears the mark. The editor selection stays a collapsed range and `isKeyboardSelectable()` stays false for the reason recorded on reference chips. Escape still does not release a claim. +- Outside a submit transaction, the machine's claim follows the document: `draft-changed` carries the leading chip's claim, so deleting or displacing the chip releases the claim and a history restore re-enters claimed. A failed command submission settles to the same rule, so it returns to claimed exactly when a chip still leads; a claim adjudicated from typed text at Enter has no chip and returns to plain. +- Node transforms keep the chip on the leading seat. A chip with a preceding sibling or inside a later block becomes its literal token, which releases the claim as typing before the old token did. The chip's own transform sees sibling insertions and moves between blocks; a paragraph transform sees a block inserted before the chip's block, so no update walks the whole document. +- Deletion takes two steps. A character deletion that reaches the chip from a collapsed caret marks it selected (`data-selected` on its host; the root's `data-composer-chip-selected` hides the caret), and the next deletion in the same direction removes it whole. Deletions are decided both on the character-deletion command and on deletion input events, because iOS applies a ranged `deleteContentBackward` without that command. A key whose dispatch produced no character deletion, a click, blur, or a caret that leaves the chip clears the mark, so platform deletion shortcuts such as macOS Ctrl+H keep it. The editor selection stays a collapsed range and `isKeyboardSelectable()` stays false for the reason recorded on reference chips. Escape still does not release a claim. - `claim-decor.ts` and the text-reference precedence rule are deleted. The `slash/input-*` events, `CommandClaim`, `InputState.claim`, and trigger sources keep their fields. `CommandClaim.token` still ends with the separator, and command arguments are still read from the clipboard text after it. diff --git a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md index 1f9648a65b..f4516aea79 100644 --- a/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md +++ b/.agents/notes/implemented/architecture/2026-10-09-composer-command-chip.zh.md @@ -14,9 +14,9 @@ Status: implemented - 节点携带 `CommandClaim`，包括其 submit 闭包。历史记录恢复节点时连同字段一起恢复，因此撤销 chip 删除会恢复认领。JSON 导出写入 token 文本；闭包没有 JSON 形式，因此 JSON 导入得到纯文本节点。 - 检测投影把 chip 计为一个 U+FFFC，触发检测不会再读到命令名。剪贴板投影是不含分隔符的 token。chip 不是引用 occurrence。`EditorProjection.command` 是位于文档开头的 chip 所携带的认领。 -- 在提交事务之外，状态机的认领跟随文档：`draft-changed` 携带行首 chip 的认领，因此删除或挤开 chip 会释放认领，历史恢复会重新进入 claimed。命令提交失败时，只有草稿未变且同一个 chip 仍在行首才回到 claimed。Enter 时从输入文本裁决出的认领没有 chip，失败后回到 plain。 -- 一个根节点 transform 让 chip 保持在行首位置。前面有兄弟节点或位于后续块中的 chip 会变回字面 token，并像在旧 token 前输入那样释放认领。 -- 删除分两步。从折叠光标出发、方向触及 chip 的字符删除先把它标记为选中（宿主元素上的 `data-selected`；根元素的 `data-composer-chip-selected` 隐藏光标），同方向的下一次删除才整体移除它。其他按键、点击、失焦或光标离开 chip 都会清除标记。编辑器选区保持为折叠的 range，`isKeyboardSelectable()` 仍为 false，原因与引用 chip 记录的相同。Escape 仍不释放认领。 +- 在提交事务之外，状态机的认领跟随文档：`draft-changed` 携带行首 chip 的认领，因此删除或挤开 chip 会释放认领，历史恢复会重新进入 claimed。命令提交失败后按同一规则结算，只要行首仍有 chip 就回到 claimed；Enter 时从输入文本裁决出的认领没有 chip，回到 plain。 +- 节点 transform 让 chip 保持在行首位置。前面有兄弟节点或位于后续块中的 chip 会变回字面 token，并像在旧 token 前输入那样释放认领。chip 自身的 transform 覆盖兄弟节点插入和跨块移动；段落 transform 覆盖在 chip 所在块之前插入块的情况，因此没有任何更新需要遍历整个文档。 +- 删除分两步。从折叠光标出发、方向触及 chip 的字符删除先把它标记为选中（宿主元素上的 `data-selected`；根元素的 `data-composer-chip-selected` 隐藏光标），同方向的下一次删除才整体移除它。字符删除命令和删除类输入事件都会参与判断，因为 iOS 处理带范围的 `deleteContentBackward` 时不经过该命令。分发过程中没有产生字符删除的按键、点击、失焦或光标离开 chip 会清除标记，因此 macOS Ctrl+H 这类平台删除快捷键会保留它。编辑器选区保持为折叠的 range，`isKeyboardSelectable()` 仍为 false，原因与引用 chip 记录的相同。Escape 仍不释放认领。 - 删除 `claim-decor.ts` 和文本引用的优先级规则。 `slash/input-*` 事件、`CommandClaim`、`InputState.claim` 与触发源的字段保持不变。`CommandClaim.token` 仍以分隔符结尾，命令参数仍从其后的剪贴板文本读取。 diff --git a/apps/web/tests/lifecycle-chrome.e2e.ts b/apps/web/tests/lifecycle-chrome.e2e.ts index 3a9ec7e07b..a0259aa131 100644 --- a/apps/web/tests/lifecycle-chrome.e2e.ts +++ b/apps/web/tests/lifecycle-chrome.e2e.ts @@ -151,6 +151,8 @@ describe('web e2e: lifecycle & chrome (workspace flow / reload / dark mode)', () await writeComposerDraft(inputPage, input, '/') await inputPage.getByRole('listbox').getByRole('option', { name: row, exact: true }).click() await expect.poll(() => input.textContent()).toBe(`${token} `) + const inputHeight = () => input.evaluate(element => element.getBoundingClientRect().height) + const claimedHeight = await inputHeight() await input.press('End') await inputPage.keyboard.insertText('这是任务') await expect.poll(() => input.textContent()).toBe(`${token} 这是任务`) @@ -159,6 +161,8 @@ describe('web e2e: lifecycle & chrome (workspace flow / reload / dark mode)', () await expect.poll(() => input.textContent()).toBe(token) await expect.poll(() => input.getAttribute('data-phase')).toBe('claimed') await expect.poll(tokenText).toBe(token) + // The chip now ends its block, where Lexical appends a managed line break; it must not add a line. + expect(await inputHeight()).toBe(claimedHeight) await input.press('Space') await expect.poll(() => input.getAttribute('data-phase')).toBe('claimed') await expect.poll(() => input.textContent()).toBe(`${token} `) diff --git a/packages/client/ui-conversation/README.i18n.yaml b/packages/client/ui-conversation/README.i18n.yaml index e456dae371..b5d9e93056 100644 --- a/packages/client/ui-conversation/README.i18n.yaml +++ b/packages/client/ui-conversation/README.i18n.yaml @@ -30,8 +30,8 @@ en: 9f4e0da7c8addf88 zh: 1a33d9981af75643 /deepseek-ai-dsh-client-ui-conversation/known-limitations-and-deferred-work: - en: 38d243630d0e4c38 - zh: a58cdc35b664020b + en: 5aba305846ca2d04 + zh: 913dbdbcfad6aa10 /deepseek-ai-dsh-client-ui-conversation/known-limitations-and-deferred-work/dev-note: en: e67f633dc6fe7773 zh: e88aeb617ffb74ea diff --git a/packages/client/ui-conversation/README.md b/packages/client/ui-conversation/README.md index 207d38343b..6c100e228b 100644 --- a/packages/client/ui-conversation/README.md +++ b/packages/client/ui-conversation/README.md @@ -167,6 +167,7 @@ None; Conversation assembly and browser input state do not alter provider-side p - **Only registered targets can render** — the shell deliberately has no implicit fallback target beyond the registered `chat` preference. - **Factory occurrences inherit their render-position Session** — `conversation.content` does not accept an independently addressed Session; that requires a separate Session-provider capability. +- **A claimed command does not survive reload** — the persisted draft keeps text and references only, so a reloaded draft shows the command as its literal token instead of a chip; Enter still adjudicates and submits it as the command. Restoring the chip needs a side-effect-free way to re-resolve the claim from the command source. diff --git a/packages/client/ui-conversation/README.zh.md b/packages/client/ui-conversation/README.zh.md index 806ad8c74c..3e9b411db3 100644 --- a/packages/client/ui-conversation/README.zh.md +++ b/packages/client/ui-conversation/README.zh.md @@ -167,6 +167,7 @@ selector 必须是 owner currency 的纯函数。非 null 返回值作为 `match - **只有已注册 target 可以渲染**——除已注册的 `chat` 偏好外，shell 刻意不提供隐式 fallback target。 - **Factory occurrence 继承渲染位置的 Session**——`conversation.content` 不接受独立寻址的 Session；该能力需要单独的 Session provider。 +- **已认领命令不会跨刷新保留**——持久化的草稿只保存文字和引用，刷新后的草稿把命令显示为字面 token 而不是 chip；Enter 仍会裁决并将其作为命令提交。恢复 chip 需要一种无副作用的方式，从命令来源重新解析认领。 diff --git a/packages/client/ui-conversation/src/client/contract/input.ts b/packages/client/ui-conversation/src/client/contract/input.ts index 416e46b463..e19bed97a2 100644 --- a/packages/client/ui-conversation/src/client/contract/input.ts +++ b/packages/client/ui-conversation/src/client/contract/input.ts @@ -321,12 +321,11 @@ export type InputEvent = | { readonly type: 'enter'; readonly mode: InputSubmitMode; readonly draft: string; readonly submission?: MessageSubmission } | { readonly type: 'adjudicated'; readonly attempt: SubmitAttempt; readonly outcome: PickOutcome } | { readonly type: 'adjudication-failed'; readonly attempt: SubmitAttempt; readonly message: string } - /** Settlement carries the live clipboard projection and leading command chip claim for claim re-entry decisions. */ + /** Settlement carries the leading command chip's claim, which the claim follows after a failure. */ | { readonly type: 'submit-settled' readonly attempt: SubmitAttempt readonly ok: boolean - readonly draft: string readonly command: CommandClaim | undefined readonly outcome?: SubmitOutcome readonly message?: string diff --git a/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx b/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx index 2b2760a36e..c434b5ced0 100644 --- a/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx +++ b/packages/client/ui-conversation/src/client/input/editor/command-chip-node.tsx @@ -10,8 +10,8 @@ import type { JSX } from 'react' import type { LexicalEditor, LexicalNode, NodeKey, SerializedLexicalNode, Spread, } from 'lexical' -import { $createTextNode, DecoratorNode, RootNode } from 'lexical' -import { $dfs } from '@lexical/utils' +import { $createTextNode, DecoratorNode, ParagraphNode } from 'lexical' +import { mergeRegister } from '@lexical/utils' import type { CommandClaim } from '../../contract/input.ts' import { CommandChip } from './CommandChip.tsx' @@ -88,8 +88,9 @@ export class CommandChipNode extends DecoratorNode { } /** - * Arrows step across the chip and Backspace/Delete remove it whole, for - * the reason documented on ReferenceChipNode.isKeyboardSelectable. + * Arrows step across the chip and the default character deletion removes + * it whole, for the reason documented on ReferenceChipNode.isKeyboardSelectable; + * registerCommandChipSelection inserts the selecting step before that deletion. */ override isKeyboardSelectable(): boolean { return false @@ -129,21 +130,28 @@ export function $isCommandChipNode(node: LexicalNode | null | undefined): node i return node instanceof CommandChipNode } +/** Replace a displaced chip with its literal token. */ +function $releaseChip(chip: CommandChipNode): void { + chip.replace($createTextNode(chip.getTextContent())) +} + /** * Keep command chips on the leading seat. A chip anywhere but the first - * child of the first block (text typed or a line break inserted before it) - * becomes its literal token, which releases the claim. The root transform - * runs whenever any attached node is dirty. + * child of the first block (text, a line break, or a block inserted before + * it) becomes its literal token, which releases the claim. Inserting a + * sibling before the chip or moving it to another block marks the chip dirty; + * a block inserted before the chip's block marks only that block dirty. * @param editor - the shell-owned editor. * @returns the unregister disposer. */ export function registerCommandChipSeat(editor: LexicalEditor): () => void { - return editor.registerNodeTransform(RootNode, (root) => { - for (const { node } of $dfs(root)) { - if ($isCommandChipNode(node) - && (node.getPreviousSibling() !== null || node.getParentOrThrow().getPreviousSibling() !== null)) { - node.replace($createTextNode(node.getTextContent())) - } - } - }) + return mergeRegister( + editor.registerNodeTransform(CommandChipNode, (chip) => { + if (chip.getPreviousSibling() !== null || chip.getParentOrThrow().getPreviousSibling() !== null) $releaseChip(chip) + }), + editor.registerNodeTransform(ParagraphNode, (block) => { + const first = block.getFirstChild() + if ($isCommandChipNode(first) && block.getPreviousSibling() !== null) $releaseChip(first) + }), + ) } diff --git a/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts b/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts index a6c33f0597..101b81617a 100644 --- a/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts +++ b/packages/client/ui-conversation/src/client/input/editor/command-chip-selection.ts @@ -1,23 +1,29 @@ /** - * Two-step command chip deletion. A character deletion (Backspace, Delete, - * or an input-event deletion) whose direction reaches the command chip from a - * collapsed caret first selects the chip; the next deletion in that direction - * removes it. Any other key, a click, blur, or a caret that no longer touches - * the chip clears the selection. The selected state is presentation on the - * chip's host and the root element; the editor selection stays a collapsed - * range beside the chip, so typing and arrows keep their plain-text behavior. + * Two-step command chip deletion. A character deletion whose direction + * reaches the command chip from a collapsed caret first selects the chip; the + * next deletion in that direction removes it. Deletions arrive as the + * character-deletion command (Backspace, Delete, and the platform shortcuts + * the engine maps to them) or as a deletion input event, which some platforms + * apply without that command. A key that produces no character deletion, a + * click, blur, or a caret that no longer touches the chip clears the + * selection. The selected state is presentation on the chip's host and the + * root element; the editor selection stays a collapsed range beside the chip, + * so typing and arrows keep their plain-text behavior. */ import type { LexicalEditor, NodeKey } from 'lexical' import { - $getAdjacentNode, $getSelection, $isRangeSelection, BLUR_COMMAND, CLICK_COMMAND, COMMAND_PRIORITY_CRITICAL, - COMMAND_PRIORITY_HIGH, DELETE_CHARACTER_COMMAND, KEY_DOWN_COMMAND, + $getAdjacentNode, $getSelection, $isRangeSelection, BEFORE_INPUT_COMMAND, BLUR_COMMAND, CLICK_COMMAND, + COMMAND_PRIORITY_CRITICAL, COMMAND_PRIORITY_HIGH, DELETE_CHARACTER_COMMAND, KEY_DOWN_COMMAND, } from 'lexical' import { mergeRegister } from '@lexical/utils' import { $isCommandChipNode } from './command-chip-node.tsx' import type { CommandChipNode } from './command-chip-node.tsx' -/** Keys that keep a selected chip: deletions and bare modifiers. */ -const RETAINING_KEYS = new Set(['Backspace', 'Delete', 'Shift', 'Control', 'Alt', 'Meta']) +/** + * Keys that never clear a selected chip by themselves: bare modifiers, and + * IME or virtual-keyboard keys whose deletion arrives later as an input event. + */ +const PASSIVE_KEYS = new Set(['Shift', 'Control', 'Alt', 'Meta', 'Unidentified', 'Process']) /** The chip a collapsed caret deletes next in one direction, or null. */ function $commandChipBesideCaret(isBackward: boolean): CommandChipNode | null { @@ -27,6 +33,13 @@ function $commandChipBesideCaret(isBackward: boolean): CommandChipNode | null { return $isCommandChipNode(node) ? node : null } +/** Deletion direction of an input event type; null for other input. */ +function deletionDirection(inputType: string): boolean | null { + if (inputType === 'deleteContentBackward') return true + if (inputType === 'deleteContentForward') return false + return null +} + /** * Register two-step deletion for the command chip. * @param editor - the shell-owned editor. @@ -34,6 +47,8 @@ function $commandChipBesideCaret(isBackward: boolean): CommandChipNode | null { */ export function registerCommandChipSelection(editor: LexicalEditor): () => void { let selected: { readonly key: NodeKey; readonly isBackward: boolean } | null = null + /** A keydown with a selected chip whose dispatch has produced no character deletion yet. */ + let keyWithoutDeletion = false const paint = (next: typeof selected): void => { if (selected !== null && selected.key !== next?.key) { editor.getElementByKey(selected.key)?.removeAttribute('data-selected') @@ -46,14 +61,41 @@ export function registerCommandChipSelection(editor: LexicalEditor): () => void if (selected !== null) paint(null) return false } + /** + * Decide one character deletion. + * @param isBackward - deletion direction. + * @param clearOnPass - whether letting the deletion run clears the selection here; + * an input event leaves it for the deletion that follows. + * @returns true when the deletion only selected the chip and must be consumed. + */ + const $decide = (isBackward: boolean, clearOnPass: boolean): boolean => { + keyWithoutDeletion = false + const chip = $commandChipBesideCaret(isBackward) + if (chip === null) return clear() + if (selected?.key === chip.getKey() && selected.isBackward === isBackward) return clearOnPass ? clear() : false + paint({ key: chip.getKey(), isBackward }) + return true + } return mergeRegister( - editor.registerCommand(DELETE_CHARACTER_COMMAND, (isBackward) => { - const chip = $commandChipBesideCaret(isBackward) - if (chip === null || (selected?.key === chip.getKey() && selected.isBackward === isBackward)) return clear() - paint({ key: chip.getKey(), isBackward }) + editor.registerCommand(BEFORE_INPUT_COMMAND, (event) => { + const isBackward = deletionDirection(event.inputType) + if (isBackward === null || !$decide(isBackward, false)) return false + event.preventDefault() return true }, COMMAND_PRIORITY_HIGH), - editor.registerCommand(KEY_DOWN_COMMAND, event => RETAINING_KEYS.has(event.key) ? false : clear(), COMMAND_PRIORITY_CRITICAL), + editor.registerCommand(DELETE_CHARACTER_COMMAND, isBackward => $decide(isBackward, true), COMMAND_PRIORITY_HIGH), + editor.registerCommand(KEY_DOWN_COMMAND, (event) => { + if (selected === null || PASSIVE_KEYS.has(event.key)) return false + // Key bindings dispatch their deletion synchronously inside this + // keydown, so the microtask sees whether the key deleted anything. + keyWithoutDeletion = true + queueMicrotask(() => { + if (!keyWithoutDeletion) return + keyWithoutDeletion = false + clear() + }) + return false + }, COMMAND_PRIORITY_CRITICAL), editor.registerCommand(CLICK_COMMAND, clear, COMMAND_PRIORITY_CRITICAL), editor.registerCommand(BLUR_COMMAND, clear, COMMAND_PRIORITY_CRITICAL), editor.registerUpdateListener(({ editorState }) => { diff --git a/packages/client/ui-conversation/src/client/input/facade.ts b/packages/client/ui-conversation/src/client/input/facade.ts index b3d7d74fd0..398c17a62a 100644 --- a/packages/client/ui-conversation/src/client/input/facade.ts +++ b/packages/client/ui-conversation/src/client/input/facade.ts @@ -847,7 +847,7 @@ export class SessionInputShell implements SessionInput { } this.dispatchRun(({ type: 'submit-settled', attempt, ok: outcome.kind === 'success', - draft: this.projection.clipboardText, command: this.projection.command, outcome, + command: this.projection.command, outcome, ...(outcome.kind === 'error' && outcome.text === undefined ? { message: 'command failed' } : {}), })) }, @@ -856,7 +856,7 @@ export class SessionInputShell implements SessionInput { const message = error instanceof Error ? error.message : String(error) this.dispatchRun(({ type: 'submit-settled', attempt, ok: false, - draft: this.projection.clipboardText, command: this.projection.command, message, + command: this.projection.command, message, })) }, ) diff --git a/packages/client/ui-conversation/src/client/input/machine.ts b/packages/client/ui-conversation/src/client/input/machine.ts index 1eec268602..b85464e5be 100644 --- a/packages/client/ui-conversation/src/client/input/machine.ts +++ b/packages/client/ui-conversation/src/client/input/machine.ts @@ -70,8 +70,8 @@ export class SubmitMachine { */ dispatch(ev: InputEvent): readonly InputEffect[] { switch (ev.type) { - case 'draft-changed': return this.onDraftChanged(ev.command) - case 'claim': return this.onClaim(ev.claim) + case 'draft-changed': return this.followCommand(ev.command) + case 'claim': return this.followCommand(ev.claim) case 'enter': return this.onEnter(ev.mode, ev.draft, ev.submission) case 'adjudicated': return this.onAdjudicated(ev.attempt, ev.outcome) case 'adjudication-failed': return this.onAdjudicationFailed(ev.attempt, ev.message) @@ -86,23 +86,17 @@ export class SubmitMachine { /** * Outside a transaction the claim follows the document's leading command * chip: deleting or displacing the chip releases it, and a history restore - * of the chip re-enters claimed. + * of the chip re-enters claimed. The claim event applies the same rule to a + * chip the editor just inserted, before its update commits; busy phases + * refuse both. */ - private onDraftChanged(command: CommandClaim | undefined): readonly InputEffect[] { + private followCommand(command: CommandClaim | undefined): readonly InputEffect[] { if (this.phase !== 'plain' && this.phase !== 'claimed') return [] this.claim = command this.phase = command === undefined ? 'plain' : 'claimed' return [] } - /** The editor inserted a command chip; busy phases refuse another claim. */ - private onClaim(claim: CommandClaim): readonly InputEffect[] { - if (this.phase !== 'plain' && this.phase !== 'claimed') return [] - this.claim = claim - this.phase = 'claimed' - return [] - } - /** Mint an attempt and controller without assigning its lifecycle owner. */ private mintAttempt(mode: InputSubmitMode, draft: string, submission?: MessageSubmission): { readonly attempt: SubmitAttempt @@ -187,7 +181,7 @@ export class SubmitMachine { return [{ type: 'notice', level: 'error', text: message }] } - /** Claimed command settlement retains the frozen transaction semantics. */ + /** Claimed command settlement; after a failure the claim follows the leading command chip again. */ private onSubmitSettled(ev: Extract): readonly InputEffect[] { const flight = this.inflight if (this.phase !== 'submitting' || flight === undefined || flight.attempt.seq !== ev.attempt.seq) return [] @@ -202,12 +196,8 @@ export class SubmitMachine { return effects } const text = ev.message ?? ev.outcome?.text - if (ev.draft === flight.attempt.draftSnapshot && this.claim !== undefined && ev.command === this.claim) { - this.phase = 'claimed' - return text === undefined ? [] : [{ type: 'notice', level: 'error', text }] - } - this.phase = 'plain' - this.claim = undefined + this.claim = ev.command + this.phase = ev.command === undefined ? 'plain' : 'claimed' return text === undefined ? [] : [{ type: 'notice', level: 'error', text }] } diff --git a/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx b/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx index d30124bf24..905b348dc8 100644 --- a/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx +++ b/packages/client/ui-conversation/tests/input-scenarios.client.spec.tsx @@ -12,7 +12,10 @@ import './control-row-dom.ts' import type { GlobalStandardProps } from '@deepseek-ai/dsh-client-ui-slots' import { afterEach, describe, expect, it, onTestFinished, vi } from 'vitest' import { act, cleanup, fireEvent, render } from '@testing-library/react' -import { $createTextNode, $getRoot, $isElementNode, DELETE_CHARACTER_COMMAND, UNDO_COMMAND } from 'lexical' +import { + $createTextNode, $getRoot, $isElementNode, BEFORE_INPUT_COMMAND, COMMAND_PRIORITY_LOW, DELETE_CHARACTER_COMMAND, + UNDO_COMMAND, +} from 'lexical' import type { SessionSnapshot } from '@deepseek-ai/dsh-api-session-controller/client' import { createSnapshotStore } from '@deepseek-ai/dsh-client-store' import { InputTriggerService } from '@deepseek-ai/dsh-client-ui-input-trigger/client' @@ -388,7 +391,7 @@ describe('scenario H: the command chip deletes whole', () => { expect(b.textarea.hasAttribute('data-composer-chip-selected')).toBe(false) }) - it('another key or typing clears a selected chip without deleting it', async () => { + it('a key that deletes nothing, or typing, clears a selected chip without deleting it', async () => { const b = await bench() b.type('/goal') await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) @@ -398,8 +401,10 @@ describe('scenario H: the command chip deletes whole', () => { b.deleteCharacter(true) expect(chip()?.hasAttribute('data-selected')).toBe(true) fireEvent.keyDown(b.textarea, { key: 'Shift' }) + await act(async () => {}) expect(chip()?.hasAttribute('data-selected')).toBe(true) fireEvent.keyDown(b.textarea, { key: 'Escape' }) + await act(async () => {}) expect(chip()?.hasAttribute('data-selected')).toBe(false) expect(b.textarea.hasAttribute('data-composer-chip-selected')).toBe(false) b.deleteCharacter(true) @@ -410,6 +415,58 @@ describe('scenario H: the command chip deletes whole', () => { expect(b.shell.snapshot.phase).toBe('claimed') }) + it('a platform deletion shortcut keeps the selection its first press made and deletes on the second', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.erase(1, 2) + const chip = (): Element | null => b.view.container.querySelector('[data-composer-command]') + // macOS maps Ctrl+H to a backward character deletion inside the same keydown. + const pressCtrlH = async (): Promise => { + act(() => { + fireEvent.keyDown(b.textarea, { key: 'h', ctrlKey: true }) + b.shell.editor.dispatchCommand(DELETE_CHARACTER_COMMAND, true) + b.shell.editor.update(() => {}, { discrete: true }) + }) + await act(async () => {}) + } + await pressCtrlH() + expect(chip()?.hasAttribute('data-selected')).toBe(true) + await pressCtrlH() + expect(chip()).toBeNull() + expect(b.shell.snapshot.phase).toBe('plain') + }) + + it('a deletion input event selects the chip first and lets the second one delete it', async () => { + const b = await bench() + b.type('/goal') + await vi.waitFor(() => { expect(b.controller.menu.getSnapshot().open).toBe(true) }) + await b.space() + b.erase(1, 2) + const chip = (): Element | null => b.view.container.querySelector('[data-composer-command]') + // Lexical's default input handling runs below this probe; iOS deletes a ranged + // deleteContentBackward there directly, so the selecting step must consume it first. + const reachedDefault = vi.fn(() => false) + onTestFinished(b.shell.editor.registerCommand(BEFORE_INPUT_COMMAND, reachedDefault, COMMAND_PRIORITY_LOW)) + const deleteByInput = (): InputEvent => { + const event = new InputEvent('beforeinput', { inputType: 'deleteContentBackward', cancelable: true }) + act(() => { + b.shell.editor.dispatchCommand(BEFORE_INPUT_COMMAND, event) + b.shell.editor.update(() => {}, { discrete: true }) + }) + return event + } + expect(deleteByInput().defaultPrevented).toBe(true) + expect(reachedDefault).not.toHaveBeenCalled() + expect(chip()?.hasAttribute('data-selected')).toBe(true) + expect(b.shell.snapshot.phase).toBe('claimed') + deleteByInput() + expect(reachedDefault).toHaveBeenCalledTimes(1) + expect(chip()).toBeNull() + expect(b.shell.snapshot.phase).toBe('plain') + }) + it('Delete before the chip selects it first and removes it on the second press', async () => { const b = await bench() b.type('/goal') diff --git a/packages/client/ui-conversation/tests/submit-machine.client.spec.ts b/packages/client/ui-conversation/tests/submit-machine.client.spec.ts index 4ecb8c2e92..4fd4df996c 100644 --- a/packages/client/ui-conversation/tests/submit-machine.client.spec.ts +++ b/packages/client/ui-conversation/tests/submit-machine.client.spec.ts @@ -221,7 +221,7 @@ describe('submit-machine: claimed lifecycle', () => { const fx = m.dispatch({ type: 'enter', mode: 'queue', draft: token.trimEnd() }) const begin = effectAt(fx, 0, 'begin-submit') expect(begin.args).toBe('') - m.dispatch({ type: 'submit-settled', attempt: begin.attempt, ok: false, draft: token.trimEnd(), command: claim }) + m.dispatch({ type: 'submit-settled', attempt: begin.attempt, ok: false, command: claim }) expect(m.state.phase).toBe('claimed') expect(m.state.claim?.name).toBe(name) }) @@ -262,7 +262,7 @@ describe('submit-machine: submitting transaction', () => { const m = new SubmitMachine() const { attempt } = enterSubmitting(m, 'goal', 'x') const fx = m.dispatch({ - type: 'submit-settled', attempt, ok: true, draft: '/goal x', command: undefined, + type: 'submit-settled', attempt, ok: true, command: undefined, outcome: { kind: 'success', text: 'goal saved' }, }) expect(effectAt(fx, 0, 'commit-draft').retainSuffixOf).toBe('/goal x') @@ -275,25 +275,26 @@ describe('submit-machine: submitting transaction', () => { const m = new SubmitMachine() const { attempt } = enterSubmitting(m, 'goal', 'x') const fx = m.dispatch({ - type: 'submit-settled', attempt, ok: false, draft: 'deviated', command: undefined, + type: 'submit-settled', attempt, ok: false, command: undefined, outcome: { kind: 'error', text: 'rejected' }, }) expect(effectAt(fx, 0, 'notice')).toMatchObject({ level: 'error', text: 'rejected' }) expect(m.state.phase).toBe('plain') }) - it('rollback with an undeviated draft and chip keeps the claim and re-enters claimed', () => { + it('a failed settlement re-enters claimed while the command chip still leads', () => { const m = new SubmitMachine() const { attempt, claim } = enterSubmitting(m, 'goal', 'x') - m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: '/goal x', command: claim, message: 'transport' }) + const fx = m.dispatch({ type: 'submit-settled', attempt, ok: false, command: claim, message: 'transport' }) + expect(effectAt(fx, 0, 'notice')).toMatchObject({ level: 'error', text: 'transport' }) expect(m.state.phase).toBe('claimed') expect(m.state.claim?.token).toBe('/goal ') }) - it('rollback with a deviated draft only notices — the newer input wins', () => { + it('a failed settlement without the command chip only notices and returns to plain', () => { const m = new SubmitMachine() - const { attempt, claim } = enterSubmitting(m, 'goal', 'x') - const fx = m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: 'rewritten', command: claim, message: 'transport' }) + const { attempt } = enterSubmitting(m, 'goal', 'x') + const fx = m.dispatch({ type: 'submit-settled', attempt, ok: false, command: undefined, message: 'transport' }) expect(effectAt(fx, 0, 'notice')).toMatchObject({ level: 'error', text: 'transport' }) expect(m.state.phase).toBe('plain') expect(m.state.claim).toBeUndefined() @@ -303,7 +304,7 @@ describe('submit-machine: submitting transaction', () => { const m = new SubmitMachine() const attempt = enterAdjudicating(m, '/goal x') m.dispatch({ type: 'adjudicated', attempt, outcome: { claim: claimOf('goal') } }) - m.dispatch({ type: 'submit-settled', attempt, ok: false, draft: '/goal x', command: undefined, message: 'nope' }) + m.dispatch({ type: 'submit-settled', attempt, ok: false, command: undefined, message: 'nope' }) expect(m.state.phase).toBe('plain') expect(m.state.claim).toBeUndefined() }) @@ -311,12 +312,12 @@ describe('submit-machine: submitting transaction', () => { it('a stale settle after rollback + resubmit is dropped (anti-backwash)', () => { const m = new SubmitMachine() const { attempt: first, claim } = enterSubmitting(m, 'goal', 'x') - m.dispatch({ type: 'submit-settled', attempt: first, ok: false, draft: '/goal x', command: claim, message: 'try again' }) + m.dispatch({ type: 'submit-settled', attempt: first, ok: false, command: claim, message: 'try again' }) const fx = m.dispatch({ type: 'enter', mode: 'queue', draft: '/goal x' }) const second = effectAt(fx, 0, 'begin-submit').attempt - expect(m.dispatch({ type: 'submit-settled', attempt: first, ok: true, draft: '/goal x', command: claim })).toEqual([]) + expect(m.dispatch({ type: 'submit-settled', attempt: first, ok: true, command: claim })).toEqual([]) expect(m.state.phase).toBe('submitting') - m.dispatch({ type: 'submit-settled', attempt: second, ok: true, draft: '/goal x', command: claim }) + m.dispatch({ type: 'submit-settled', attempt: second, ok: true, command: claim }) expect(m.state.phase).toBe('plain') }) @@ -325,7 +326,7 @@ describe('submit-machine: submitting transaction', () => { const { attempt } = enterSubmitting(m, 'goal', 'x') m.dispatch({ type: 'release' }) expect(attempt.signal.aborted).toBe(true) - expect(m.dispatch({ type: 'submit-settled', attempt, ok: true, draft: '', command: undefined })).toEqual([]) + expect(m.dispatch({ type: 'submit-settled', attempt, ok: true, command: undefined })).toEqual([]) expect(m.state.phase).toBe('plain') }) @@ -346,7 +347,7 @@ describe('submit-machine: per-session isolation', () => { const { attempt } = enterSubmitting(a, 'goal', 'x') const fx = b.dispatch({ type: 'enter', mode: 'queue', draft: 'hello' }) expect(effectAt(fx, 0, 'default-sink').draft).toBe('hello') - a.dispatch({ type: 'submit-settled', attempt, ok: true, draft: '/goal x', command: undefined }) + a.dispatch({ type: 'submit-settled', attempt, ok: true, command: undefined }) expect(a.state.phase).toBe('plain') expect(b.state.phase).toBe('plain') }) From ea5d070e373f8d813dc392454810f8eca2b2149b Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 13:53:20 +0800 Subject: [PATCH 008/162] test(subagent): refresh model tools in merged snapshot scenarios --- .../tool-schemas.expected.json | 45 ++++++++++++++++++- .../system-prompt.expected.md | 24 +++++++++- .../session.v4.jsonl | 6 +-- .../ptc-persistent/system-prompt.expected.md | 24 +++++++++- .../system-prompt.expected.md | 24 +++++++++- 5 files changed, 112 insertions(+), 11 deletions(-) diff --git a/snapshots/session/office-progressive-native/tool-schemas.expected.json b/snapshots/session/office-progressive-native/tool-schemas.expected.json index 79d6891abc..7269632ffd 100644 --- a/snapshots/session/office-progressive-native/tool-schemas.expected.json +++ b/snapshots/session/office-progressive-native/tool-schemas.expected.json @@ -215,6 +215,23 @@ } } }, + { + "name": "list_subagent_models", + "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "parameters": { + "type": "object", + "properties": { + "provider": { + "type": "string", + "description": "Registered LLM provider id. Omit to list providers." + }, + "model": { + "type": "string", + "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + } + } + } + }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -294,7 +311,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", "parameters": { "type": "object", "properties": { @@ -309,6 +326,18 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -319,7 +348,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", "parameters": { "type": "object", "properties": { @@ -334,6 +363,18 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." + }, + "provider": { + "type": "string", + "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + }, + "model": { + "type": "string", + "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + }, + "reasoning_effort": { + "type": "string", + "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/office-progressive-ptc/system-prompt.expected.md b/snapshots/session/office-progressive-ptc/system-prompt.expected.md index 9d427282d8..e56a80173c 100644 --- a/snapshots/session/office-progressive-ptc/system-prompt.expected.md +++ b/snapshots/session/office-progressive-ptc/system-prompt.expected.md @@ -119,6 +119,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -145,7 +152,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -153,8 +160,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -162,6 +175,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -350,6 +369,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl b/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl index 843a75e88d..6e26da9e41 100644 --- a/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl +++ b/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"moonshotai","model":"kimi-k3","contextWindow":1048576,"systemPromptUpdate":"in-history"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"moonshotai","model":"kimi-k3"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790149644504,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644504,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790149644504,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -18,14 +18,14 @@ {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"moonshotai","model":"kimi-k3"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1790149644534,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644534,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790149644534,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790149644535,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:9}}"}},"sourceEventSeqs":[22],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:10}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"moonshotai","model":"kimi-k3"},"id":"{{message:11}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790149644557,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790149644557,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790149644557,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/session/ptc-persistent/system-prompt.expected.md b/snapshots/session/ptc-persistent/system-prompt.expected.md index 2801645046..835372b136 100644 --- a/snapshots/session/ptc-persistent/system-prompt.expected.md +++ b/snapshots/session/ptc-persistent/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { @@ -411,6 +430,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-policy-change/system-prompt.expected.md b/snapshots/session/ptc-policy-change/system-prompt.expected.md index 19204006b3..8308b029e0 100644 --- a/snapshots/session/ptc-policy-change/system-prompt.expected.md +++ b/snapshots/session/ptc-policy-change/system-prompt.expected.md @@ -140,6 +140,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -166,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -174,8 +181,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -183,6 +196,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -409,6 +428,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; From ae19c6df8ee9beecd04a21a61a23e0c85444d5cb Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 14:02:48 +0800 Subject: [PATCH 009/162] test(web): align merged snapshots with shared subagent tools --- .../collapse-timing/next-message.expected.md | 2 +- .../web/collapse-timing/verbose.expected.md | 2 +- .../system-prompt.expected.md | 24 +++++++++++++++++-- .../system-prompt.expected.md | 24 +++++++++++++++++-- 4 files changed, 46 insertions(+), 6 deletions(-) diff --git a/snapshots/web/collapse-timing/next-message.expected.md b/snapshots/web/collapse-timing/next-message.expected.md index 5a7c69c657..8968e9991c 100644 --- a/snapshots/web/collapse-timing/next-message.expected.md +++ b/snapshots/web/collapse-timing/next-message.expected.md @@ -11,7 +11,7 @@ - button "Copy" - status: Completed - button "Completed in {{duration}}" [expanded] -- button "Tools updated 31 added" +- button "Tools updated 32 added" - button "Think Both files have been read. a.txt contains \"alpha\" and b.txt contains \"beta\". I'll now reply with DONE as instructed." - paragraph: DONE - button "Copy" diff --git a/snapshots/web/collapse-timing/verbose.expected.md b/snapshots/web/collapse-timing/verbose.expected.md index b747e28130..272ff8345d 100644 --- a/snapshots/web/collapse-timing/verbose.expected.md +++ b/snapshots/web/collapse-timing/verbose.expected.md @@ -19,7 +19,7 @@ - button "Copy" - status: Completed - button "Completed in {{duration}}" [disabled] [expanded] -- button "Tools updated 31 added" +- button "Tools updated 32 added" - button "Think Both files have been read. a.txt contains \"alpha\" and b.txt contains \"beta\". I'll now reply with DONE as instructed." - paragraph: DONE - button "Copy" diff --git a/snapshots/web/official-subagent-bundle/system-prompt.expected.md b/snapshots/web/official-subagent-bundle/system-prompt.expected.md index 9e8e56f62b..8908ece5a4 100644 --- a/snapshots/web/official-subagent-bundle/system-prompt.expected.md +++ b/snapshots/web/official-subagent-bundle/system-prompt.expected.md @@ -158,6 +158,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied. */ present: { /** Usually the 1-2 most important deliverables; at most 4 per call. */ @@ -281,7 +288,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -289,6 +296,12 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. */ subagent_claude_code: { @@ -299,7 +312,7 @@ interface ToolArgsMap { /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -307,6 +320,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -509,6 +528,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; present: { turn: number; files: { diff --git a/snapshots/web/optional-string-editor/system-prompt.expected.md b/snapshots/web/optional-string-editor/system-prompt.expected.md index f5e4347e82..00fa427842 100644 --- a/snapshots/web/optional-string-editor/system-prompt.expected.md +++ b/snapshots/web/optional-string-editor/system-prompt.expected.md @@ -158,6 +158,13 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; + /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + list_subagent_models: { + /** Registered LLM provider id. Omit to list providers. */ + provider?: string; + /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + model?: string; + } & Record; /** Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied. */ present: { /** Usually the 1-2 most important deliverables; at most 4 per call. */ @@ -298,7 +305,7 @@ interface ToolArgsMap { /** Optional parameter of `view` command when `path` points to a file. If omitted or null, the full file is shown. If provided, the file will be shown in the indicated line number range, e.g. [11, 12] will show lines 11 and 12. Indexing at 1 to start. Setting `[start_line, -1]` shows all lines from `start_line` to the end of the file. */ view_range?: number[] | null; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -306,8 +313,14 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -315,6 +328,12 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; + /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + provider?: string; + /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + model?: string; + /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { @@ -517,6 +536,7 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; + list_subagent_models: string; present: { turn: number; files: { From 41ea4eb6def8c65706e19edb4c1cfe82642ae053 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 15:00:27 +0800 Subject: [PATCH 010/162] fix(subagent): report model discovery errors per route --- ...authorized-subagent-model-routes.i18n.yaml | 4 +- ...4-user-authorized-subagent-model-routes.md | 2 +- ...ser-authorized-subagent-model-routes.zh.md | 2 +- .../subagent/tool-subagent/README.i18n.yaml | 4 +- packages/subagent/tool-subagent/README.md | 2 +- packages/subagent/tool-subagent/README.zh.md | 2 +- .../subagent/tool-subagent/src/list-models.ts | 15 ++++- .../tests/fixtures/model-discovery.mjs | 31 ++++++++++ .../tool-subagent/tests/list-models.spec.ts | 56 ++++++++++++++++++- .../runtime.cordis.yml | 3 + .../session.v4.jsonl | 23 ++++++++ .../snapshot.yml | 7 +++ 12 files changed, 139 insertions(+), 12 deletions(-) create mode 100644 packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs create mode 100644 snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml create mode 100644 snapshots/session/subagent-model-discovery-errors/session.v4.jsonl create mode 100644 snapshots/session/subagent-model-discovery-errors/snapshot.yml diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index c5c1287e45..d9816249cd 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -15,8 +15,8 @@ en: e41c17226190ea8b zh: d3b1395fafe75f01 /agent-note-user-authorized-subagent-model-routes/consequences: - en: 29ca812656549a01 - zh: aacd814d7d649dbf + en: e57b48c7967a316d + zh: 085fe2e57019c01c /agent-note-user-authorized-subagent-model-routes/related-decisions: en: 17fdfabedb673935 zh: 496f4f85622d7d15 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index 3e2ab42e5a..e9043a8ddb 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -33,7 +33,7 @@ Model selection has no unrestricted static mode. The default-off Host setting au ## Consequences - New adapter registrations and newly advertised models do not expand user authorization. -- Adapter removal or exact-model resolution failure can make discovery unavailable without deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. +- Adapter removal can make provider discovery unavailable; provider model listings report exact-model resolution failures per item without discarding other results or deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. - The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. - The policy event is log-only and is appended while an Agent is composed, before either SDK begins its run subscription. Shipped SDK profiles do not enable this Web-owned preference, so the event changes neither SDK's expected notifications or persisted-session output; package restore tests own its durable projection instead of fabricating an SDK composition solely to emit it. - Unit coverage pins settings validation, malformed durable values, Session sampling and inheritance, shared definitions, parent-route defaults, discovery restrictions, executor denial, and the settings UI’s staged writes and revision checks. The assembled Web scenario pins the real settings document and Plugins card flow. diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index d90bca9358..c2121c1b1c 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -33,7 +33,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` ## Consequences - 新适配器注册和新公布模型不会扩大用户授权。 -- 适配器移除或精确模型解析失败可能使发现不可用，但不会删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 +- 适配器移除可能使提供方发现不可用；提供方模型列表逐项报告精确模型解析失败，不会丢弃其他结果或删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 - 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。 - 策略事件仅存在于日志，并在 Agent 组合期间、两套 SDK 开始订阅运行前追加。随附 SDK profile 不启用这项 Web 自有偏好，因此该事件不会改变任一 SDK 的预期通知或持久 Session 输出；其持久投影由包级恢复测试负责，不会为了发出该事件而虚构 SDK 组合。 - 单元覆盖固定设置校验、异常持久值、Session 取样与继承、共享定义、父级路由默认值、发现限制、执行器拒绝，以及设置 UI 的暂存写入与 revision 检查。组装 Web 场景固定真实设置文档与 Plugins 设置卡流程。 diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index 41e61ad7ac..adbba294ab 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -66,8 +66,8 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: 14c12f23474ee1bf - zh: bd803212b3b0ba82 + en: 5bfd2012bcf18ccf + zh: d93155da44a977e1 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: en: cfdfe3b1106c3b57 zh: 3f421f2901d6f7ab diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index e407e5f7ae..9a6a9bdeeb 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -143,7 +143,7 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. Discovery resolves only allowed routes, including exact model ids absent from advisory catalogs; unauthorized provider/model lookups fail before adapter access. Execution independently enforces the same policy. +The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. Discovery resolves only allowed routes, including exact model ids absent from advisory catalogs; unauthorized provider/model lookups fail before adapter access. Provider listings retain resolved models and report lookup errors beside each failed model; exact-model queries and cancellation still reject the call. Execution independently enforces the same policy. #### Token effect diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index ea0a06a3ba..55790c1c95 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -143,7 +143,7 @@ kind: "package-reference" #### 模型看到什么 -共享的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。发现只解析允许的路由，包括建议性目录未列出的精确模型 ID；未授权的提供方／模型查询会在访问适配器前失败。执行过程独立强制同一策略。 +共享的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。发现只解析允许的路由，包括建议性目录未列出的精确模型 ID；未授权的提供方／模型查询会在访问适配器前失败。提供方列表保留成功解析的模型，并在每个失败模型旁显示查询错误；精确模型查询失败和取消仍会使调用被拒绝。执行过程独立强制同一策略。 #### Token 影响 diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index c642c4a743..72a1bac263 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -63,8 +63,18 @@ async function listSubagentModels( } const provider = registeredProvider(llm, policy, request.provider) if (request.model === undefined) { - const models = await Promise.all(allowedRoutes.map(route => llm.resolveModelInfo(provider.id, route.model, signal))) - return models.map(model => modelLine(provider.id, model)).join('\n') + const lines = await Promise.all(allowedRoutes.map(async (route) => { + let model + try { + model = await llm.resolveModelInfo(provider.id, route.model, signal) + } catch (error) { + signal.throwIfAborted() + return `${provider.id}/${route.model} — Error: ${error instanceof Error ? error.message : String(error)}` + } + return modelLine(provider.id, model) + })) + signal.throwIfAborted() + return lines.join('\n') } if (request.model.length === 0) throw new Error('`model` must be non-empty') if (!allowedRoutes.some(route => route.model === request.model)) { @@ -80,6 +90,7 @@ async function listSubagentModels( /** * Register `list_subagent_models` in the owning composition. + * Provider listings report lookup failures per model; cancellation rejects the call. * @param ctx - Context whose tool registry owns the fixed discovery definition. * @param policyFor - Resolve the calling Session’s authorized routes. */ diff --git a/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs b/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs new file mode 100644 index 0000000000..668d0aba5d --- /dev/null +++ b/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs @@ -0,0 +1,31 @@ +/** Deterministic metadata lookups for the model-discovery transcript. */ +import { LlmAdapter } from '@deepseek-ai/dsh-llm' + +class DiscoveryAdapter extends LlmAdapter { + async resolveModel(provider, model) { + if (model === 'removed') throw new Error('model was removed') + return { provider, id: model, name: 'Healthy' } + } + + async *stream() { + throw new Error('metadata-only fixture does not serve model requests') + } +} + +export const name = 'discovery-fixture' +export const inject = ['llm'] + +export function apply(ctx) { + ctx.llm.registerAdapter(['discovery-fixture'], new DiscoveryAdapter()) + ctx.on('agent/pre-step', async ({ agent, turn, step }, next) => { + if (turn === 1 && step === 1) { + agent.session.append('subagent/model-selection-policy', { + allowedModels: [ + { provider: 'discovery-fixture', model: 'removed' }, + { provider: 'discovery-fixture', model: 'healthy' }, + ], + }) + } + return next() + }) +} diff --git a/packages/subagent/tool-subagent/tests/list-models.spec.ts b/packages/subagent/tool-subagent/tests/list-models.spec.ts index 43883b48f7..e6d2fd811d 100644 --- a/packages/subagent/tool-subagent/tests/list-models.spec.ts +++ b/packages/subagent/tool-subagent/tests/list-models.spec.ts @@ -90,9 +90,9 @@ async function setupAllowedListTool() { let counter = 0 -function call(ctx: Context, args: unknown) { +function call(ctx: Context, args: unknown, signal = testToolSignal) { return ctx.tools.execute({ - signal: testToolSignal, + signal, callId: ToolCallId(`list-models-${++counter}`), name: 'list_subagent_models', arguments: args, @@ -224,6 +224,58 @@ describe('list_subagent_models', () => { expect(text(result)).toBe('alpha/fast — Fast: Focused work.') }) + it.each([false, true])('reports each failed model alongside resolved models (all failed: %s)', async (allFailed) => { + const ctx = await setupListTool([ + { provider: 'alpha', model: 'removed' }, + ...allFailed ? [] : [{ provider: 'alpha', model: 'plain' }], + { provider: 'alpha', model: 'unavailable' }, + ]) + try { + const adapter = new CatalogAdapter() + vi.spyOn(adapter, 'resolveModel').mockImplementation(async (provider, model) => { + if (model === 'removed') throw new Error('model was removed') + if (model === 'unavailable') return Promise.reject('metadata unavailable') + return { provider, id: model, name: 'Plain' } + }) + ctx.llm.registerAdapter(['alpha'], adapter) + + const result = await call(ctx, { provider: 'alpha' }) + expect(result.isError).toBe(false) + expect(text(result)).toBe([ + 'alpha/removed — Error: model was removed', + ...allFailed ? [] : ['alpha/plain — Plain'], + 'alpha/unavailable — Error: metadata unavailable', + ].join('\n')) + + const exact = await call(ctx, { provider: 'alpha', model: 'removed' }) + expect(exact.isError).toBe(true) + expect(text(exact)).toContain('model was removed') + } finally { + await ctx.fiber.dispose() + } + }) + + it.each([false, true])('does not render cancellation as a model error (adapter rejects: %s)', async (rejects) => { + const ctx = await setupListTool([{ provider: 'alpha', model: 'plain' }]) + try { + const controller = new AbortController() + const adapter = new CatalogAdapter() + vi.spyOn(adapter, 'resolveModel').mockImplementation(async (provider, model) => { + controller.abort(new Error('discovery cancelled')) + if (rejects) throw new Error('lookup interrupted') + return { provider, id: model, name: 'Plain' } + }) + ctx.llm.registerAdapter(['alpha'], adapter) + + const result = await call(ctx, { provider: 'alpha' }, controller.signal) + expect(result.isError).toBe(true) + expect(text(result)).toContain('discovery cancelled') + expect(text(result)).not.toContain('alpha/plain —') + } finally { + await ctx.fiber.dispose() + } + }) + it('inspects exact-model efforts, descriptions, and defaults', async () => { const ctx = await setupListTool() ctx.llm.registerAdapter(['alpha', 'secret'], new CatalogAdapter()) diff --git a/snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml b/snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml new file mode 100644 index 0000000000..3b8eed677f --- /dev/null +++ b/snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml @@ -0,0 +1,3 @@ +- insert: + - id: discovery-fixture + name: '../../../packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs' diff --git a/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl b/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl new file mode 100644 index 0000000000..869406b413 --- /dev/null +++ b/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl @@ -0,0 +1,23 @@ +{"type":"session","version":4,"id":"{{session:1}}","createdAt":1783352044766,"cwd":"{{cwd}}","isSeeded":false,"delegationDepth":0} +{"type":"permission/preset","data":{"preset":"danger-full-access"}} +{"type":"sandbox/mode","data":{"mode":"danger-full-access"}} +{"type":"approval/policy","data":{"policy":"never"}} +{"type":"agent/inbox/spliced","data":{"target":"next-turn","start":0,"inserted":[{"content":[{"type":"text","text":"List the models available for subagents from discovery-fixture, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"}]}} +{"type":"turn/start","data":{"turn":1}} +{"type":"agent/inbox/spliced","data":{"target":"next-turn","start":0,"removedCount":1,"inserted":[]}} +{"type":"subagent/model-selection-policy","data":{"allowedModels":[{"provider":"discovery-fixture","model":"removed"},{"provider":"discovery-fixture","model":"healthy"}]}} +{"type":"step/start","data":{"turn":1,"step":1}} +{"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} +{"type":"user/message","data":{"content":[{"type":"text","text":"List the models available for subagents from discovery-fixture, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} +{"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":"{{tools}}"},"reason":"initial"}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} +{"type":"session/title","data":{"title":"List the models available for","messageSeqs":[9],"source":{"kind":"fallback"}}} +{"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":100,"outputTokens":10},"stream":[{"type":"chunk","time":1791788384782,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"chunk","time":1791788384782,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}}},{"type":"chunk","time":1791788384782,"chunk":{"type":"usage","usage":{"inputTokens":100,"outputTokens":10}}},{"type":"chunk","time":1791788384782,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} +{"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}} +{"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_discovery"},"toolCallId":"call_discovery","content":[{"type":"text","text":"discovery-fixture/removed — Error: model was removed\ndiscovery-fixture/healthy — Healthy"}],"isError":false,"id":"{{message:5}}"}},"sourceEventSeqs":[15],"surfaceOp":"append"} +{"type":"step/end","data":{"turn":1,"step":1}} +{"type":"step/start","data":{"turn":1,"step":2}} +{"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:6}}"},"usage":{"inputTokens":100,"outputTokens":10},"stream":[{"type":"chunk","time":1791788384799,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"chunk","time":1791788384800,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1791788384800,"chunk":{"type":"usage","usage":{"inputTokens":100,"outputTokens":10}}},{"type":"chunk","time":1791788384800,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} +{"type":"step/end","data":{"turn":1,"step":2}} +{"type":"turn/end","data":{"turn":1,"reason":{"kind":"completed"}}} diff --git a/snapshots/session/subagent-model-discovery-errors/snapshot.yml b/snapshots/session/subagent-model-discovery-errors/snapshot.yml new file mode 100644 index 0000000000..322ac5fd4f --- /dev/null +++ b/snapshots/session/subagent-model-discovery-errors/snapshot.yml @@ -0,0 +1,7 @@ +version: 1 +scenario: subagent-model-discovery-errors +profile: headless +composition: default +recording: authored +header: + class: default From 080b921b5b9e7af423daf2638fdddca4b0049f11 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 15:07:11 +0800 Subject: [PATCH 011/162] docs(subagent): clarify exact-model lookup failures --- packages/subagent/tool-subagent/README.i18n.yaml | 2 +- packages/subagent/tool-subagent/README.md | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index adbba294ab..a27bf9f046 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -66,7 +66,7 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: 5bfd2012bcf18ccf + en: b2bafbccb0e10406 zh: d93155da44a977e1 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: en: cfdfe3b1106c3b57 diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index 9a6a9bdeeb..a1a2527afb 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -143,7 +143,7 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. Discovery resolves only allowed routes, including exact model ids absent from advisory catalogs; unauthorized provider/model lookups fail before adapter access. Provider listings retain resolved models and report lookup errors beside each failed model; exact-model queries and cancellation still reject the call. Execution independently enforces the same policy. +The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. Discovery resolves only allowed routes, including exact model ids absent from advisory catalogs; unauthorized provider/model lookups fail before adapter access. Provider listings retain resolved models and report lookup errors beside each failed model; exact-model lookup failures and cancellation still reject the call. Execution independently enforces the same policy. #### Token effect From 0948cd9809cba95464eaedc93b186ef5a41d2810 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 15:14:22 +0800 Subject: [PATCH 012/162] test(subagent): document non-Error adapter rejection fixture --- packages/subagent/tool-subagent/tests/list-models.spec.ts | 1 + 1 file changed, 1 insertion(+) diff --git a/packages/subagent/tool-subagent/tests/list-models.spec.ts b/packages/subagent/tool-subagent/tests/list-models.spec.ts index e6d2fd811d..d5b7642660 100644 --- a/packages/subagent/tool-subagent/tests/list-models.spec.ts +++ b/packages/subagent/tool-subagent/tests/list-models.spec.ts @@ -234,6 +234,7 @@ describe('list_subagent_models', () => { const adapter = new CatalogAdapter() vi.spyOn(adapter, 'resolveModel').mockImplementation(async (provider, model) => { if (model === 'removed') throw new Error('model was removed') + // oxlint-disable-next-line typescript/prefer-promise-reject-errors -- Exercise non-Error rejections from an external adapter. if (model === 'unavailable') return Promise.reject('metadata unavailable') return { provider, id: model, name: 'Plain' } }) From 7609189262a8915f0489fc08c846085d5d3ea5cf Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 15:27:02 +0800 Subject: [PATCH 013/162] fix(subagent): intersect model discovery with authorized catalog --- ...authorized-subagent-model-routes.i18n.yaml | 8 +- ...4-user-authorized-subagent-model-routes.md | 4 +- ...ser-authorized-subagent-model-routes.zh.md | 4 +- .../subagent/tool-subagent/README.i18n.yaml | 4 +- packages/subagent/tool-subagent/README.md | 2 +- packages/subagent/tool-subagent/README.zh.md | 2 +- .../subagent/tool-subagent/src/list-models.ts | 49 +++++------ .../tests/fixtures/model-discovery.mjs | 7 ++ .../tool-subagent/tests/list-models.spec.ts | 84 +++++++++++++------ .../session.v4.jsonl | 2 +- 10 files changed, 102 insertions(+), 64 deletions(-) diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index d9816249cd..cd7f0e6dca 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -9,14 +9,14 @@ en: 3238dfb7db7dfdd8 zh: ec3bdd9a9da0a321 /agent-note-user-authorized-subagent-model-routes/decision: - en: 499721e05ad15d9c - zh: 2bcf5ade8910da8e + en: 47b020441723cc06 + zh: 8394c32b711a0b30 /agent-note-user-authorized-subagent-model-routes/alternatives-considered: en: e41c17226190ea8b zh: d3b1395fafe75f01 /agent-note-user-authorized-subagent-model-routes/consequences: - en: e57b48c7967a316d - zh: 085fe2e57019c01c + en: 1749cfbfda849692 + zh: 9925e638cd72fba6 /agent-note-user-authorized-subagent-model-routes/related-decisions: en: 17fdfabedb673935 zh: 496f4f85622d7d15 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index e9043a8ddb..aa0de7fad7 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -14,7 +14,7 @@ The Host-owned `subagent-model-selection` settings section stores an explicit `e A newly composed top-level Session snapshots the route list in `subagent/model-selection-policy` when the setting is enabled, before its first model request. Event presence means selection was enabled; the event does not store the global switch. Child Sessions inherit that exact list from their live parent, and resumed Sessions use the recorded event instead of current settings. Settings changes therefore affect only subsequently composed top-level Sessions. Without the event, including an explicitly empty restored Session, explicit selection remains restricted to the current parent route. -The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and resolves their allowed exact model ids, including ids absent from advisory catalogs. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts. Without a recorded policy, discovery uses the parent Agent’s current effective route. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. +The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and intersects their advertised model catalogs with the recorded route list. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts, including for ids absent from the catalog. Without a recorded policy, discovery resolves only the parent Agent’s current effective route, even if the catalog omits it. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. The primary spawn tool captures the setting once, while all capable delegation tools and shared discovery read the same Session policy. Fork defaults preserve the parent route, and its description warns about prefix recomputation when that route changes. @@ -33,7 +33,7 @@ Model selection has no unrestricted static mode. The default-off Host setting au ## Consequences - New adapter registrations and newly advertised models do not expand user authorization. -- Adapter removal can make provider discovery unavailable; provider model listings report exact-model resolution failures per item without discarding other results or deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. +- Adapter removal or catalog lookup failure can make provider discovery unavailable without deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. - The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. - The policy event is log-only and is appended while an Agent is composed, before either SDK begins its run subscription. Shipped SDK profiles do not enable this Web-owned preference, so the event changes neither SDK's expected notifications or persisted-session output; package restore tests own its durable projection instead of fabricating an SDK composition solely to emit it. - Unit coverage pins settings validation, malformed durable values, Session sampling and inheritance, shared definitions, parent-route defaults, discovery restrictions, executor denial, and the settings UI’s staged writes and revision checks. The assembled Web scenario pins the real settings document and Plugins card flow. diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index c2121c1b1c..c157cec8a7 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -14,7 +14,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` 设置启用时，新组合的顶层 Session 在首个模型请求前将路由列表记录为 `subagent/model-selection-policy`。事件存在表示选择已启用；事件不保存全局开关。子 Session 从在线父级继承同一精确列表，恢复的 Session 使用已记录事件而非当前设置。因此设置变更只影响之后组合的顶层 Session。没有该事件时，包括显式为空的恢复 Session，显式选择仍仅限当前父级路由。 -固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并解析允许的精确模型 ID，包括建议性目录未列出的 ID。精确查询先要求授权，再解析模型元数据及全部支持的推理强度。没有已记录策略时，发现使用父 Agent 当前生效的路由。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 +固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并取其公布模型目录与已记录路由列表的交集。精确查询先要求授权，再解析模型元数据及全部支持的推理强度，也允许目录未列出的 ID。没有已记录策略时，发现只解析父 Agent 当前生效的路由，即使目录未列出它。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。主 spawn 工具一次性记录设置，所有支持选择的委派工具和共享发现均读取同一 Session 策略。fork 默认保留父级路由，其描述会提示改变路由时的前缀重算成本。 @@ -33,7 +33,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` ## Consequences - 新适配器注册和新公布模型不会扩大用户授权。 -- 适配器移除可能使提供方发现不可用；提供方模型列表逐项报告精确模型解析失败，不会丢弃其他结果或删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 +- 适配器移除或目录查询失败可能使提供方发现不可用，但不会删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 - 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。 - 策略事件仅存在于日志，并在 Agent 组合期间、两套 SDK 开始订阅运行前追加。随附 SDK profile 不启用这项 Web 自有偏好，因此该事件不会改变任一 SDK 的预期通知或持久 Session 输出；其持久投影由包级恢复测试负责，不会为了发出该事件而虚构 SDK 组合。 - 单元覆盖固定设置校验、异常持久值、Session 取样与继承、共享定义、父级路由默认值、发现限制、执行器拒绝，以及设置 UI 的暂存写入与 revision 检查。组装 Web 场景固定真实设置文档与 Plugins 设置卡流程。 diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index a27bf9f046..e59a887fdd 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -66,8 +66,8 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: b2bafbccb0e10406 - zh: d93155da44a977e1 + en: be3c5b647252ddc7 + zh: efecd46ed7dd9a95 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: en: cfdfe3b1106c3b57 zh: 3f421f2901d6f7ab diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index a1a2527afb..96dcfabfe6 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -143,7 +143,7 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. Discovery resolves only allowed routes, including exact model ids absent from advisory catalogs; unauthorized provider/model lookups fail before adapter access. Provider listings retain resolved models and report lookup errors beside each failed model; exact-model lookup failures and cancellation still reject the call. Execution independently enforces the same policy. +The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. #### Token effect diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index 55790c1c95..0bf8ee903f 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -143,7 +143,7 @@ kind: "package-reference" #### 模型看到什么 -共享的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。发现只解析允许的路由，包括建议性目录未列出的精确模型 ID；未授权的提供方／模型查询会在访问适配器前失败。提供方列表保留成功解析的模型，并在每个失败模型旁显示查询错误；精确模型查询失败和取消仍会使调用被拒绝。执行过程独立强制同一策略。 +共享的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 #### Token 影响 diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index 72a1bac263..950cf09e72 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -38,10 +38,20 @@ function modelLine(provider: string, model: { id: string; name: string; descript /** Read the requested provider, allowed models, or exact-model efforts. */ async function listSubagentModels( ctx: Context, - policy: ModelSelectionPolicy, + policy: ModelSelectionPolicy | undefined, request: ListSubagentModelsRequest, - signal: AbortSignal, + exec: ToolRunContext, ): Promise { + const selectionEnabled = policy !== undefined + if (policy === undefined) { + if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') + const { provider, model } = parentAgentOptionsForDelegation(exec.agent) + if (provider === undefined || model === undefined) { + throw new Error('cannot discover child LLM routes without a parent provider and model') + } + policy = { routes: [{ provider, model }] } + } + const { signal } = exec const llm = ctx.get('llm') if (llm === undefined) { throw new Error('cannot discover child LLM routes because the `llm` service is unavailable') @@ -58,23 +68,19 @@ async function listSubagentModels( } if (request.provider.length === 0) throw new Error('`provider` must be non-empty') const allowedRoutes = policy.routes.filter(route => route.provider === request.provider) - if (allowedRoutes.length === 0) { + const [firstRoute] = allowedRoutes + if (firstRoute === undefined) { throw new Error(`LLM provider "${request.provider}" is not allowed for this Session`) } const provider = registeredProvider(llm, policy, request.provider) if (request.model === undefined) { - const lines = await Promise.all(allowedRoutes.map(async (route) => { - let model - try { - model = await llm.resolveModelInfo(provider.id, route.model, signal) - } catch (error) { - signal.throwIfAborted() - return `${provider.id}/${route.model} — Error: ${error instanceof Error ? error.message : String(error)}` - } - return modelLine(provider.id, model) - })) + const models = selectionEnabled + ? (await llm.listModels(provider.id)).filter(model => allowedRoutes.some(route => route.model === model.id)) + : [await llm.resolveModelInfo(provider.id, firstRoute.model, signal)] signal.throwIfAborted() - return lines.join('\n') + return models.length === 0 + ? `(no advertised models for ${provider.id})` + : models.map(model => modelLine(provider.id, model)).join('\n') } if (request.model.length === 0) throw new Error('`model` must be non-empty') if (!allowedRoutes.some(route => route.model === request.model)) { @@ -90,11 +96,11 @@ async function listSubagentModels( /** * Register `list_subagent_models` in the owning composition. - * Provider listings report lookup failures per model; cancellation rejects the call. + * Provider listings intersect the catalog with authorization, or resolve only the parent model when selection is disabled. * @param ctx - Context whose tool registry owns the fixed discovery definition. - * @param policyFor - Resolve the calling Session’s authorized routes. + * @param policyFor - Read the calling Session’s authorized routes; undefined restricts discovery to its parent model. */ -export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolRunContext) => ModelSelectionPolicy): void { +export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolRunContext) => ModelSelectionPolicy | undefined): void { ctx.tools.register(defineTool({ name: 'list_subagent_models', description: @@ -118,7 +124,7 @@ export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolR render: (_args, result) => [{ type: 'text', text: result }], }, execute(args, exec) { - return listSubagentModels(ctx, policyFor(exec), args, exec.signal) + return listSubagentModels(ctx, policyFor(exec), args, exec) }, })) } @@ -137,11 +143,6 @@ export function apply(ctx: Context): void { registerListSubagentModels(ctx, (exec) => { if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') const routes = subagentModelSelectionPolicy(ctx.sessionProjections, exec.agent.session) - if (routes !== undefined) return { routes } - const { provider, model } = parentAgentOptionsForDelegation(exec.agent) - if (provider === undefined || model === undefined) { - throw new Error('cannot discover child LLM routes without a parent provider and model') - } - return { routes: [{ provider, model }] } + return routes === undefined ? undefined : { routes } }) } diff --git a/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs b/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs index 668d0aba5d..871ddc0ffc 100644 --- a/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs +++ b/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs @@ -2,6 +2,13 @@ import { LlmAdapter } from '@deepseek-ai/dsh-llm' class DiscoveryAdapter extends LlmAdapter { + async listModels(provider) { + return [ + { provider, id: 'healthy', name: 'Healthy' }, + { provider, id: 'unauthorized', name: 'Unauthorized' }, + ] + } + async resolveModel(provider, model) { if (model === 'removed') throw new Error('model was removed') return { provider, id: model, name: 'Healthy' } diff --git a/packages/subagent/tool-subagent/tests/list-models.spec.ts b/packages/subagent/tool-subagent/tests/list-models.spec.ts index d5b7642660..00dbca7238 100644 --- a/packages/subagent/tool-subagent/tests/list-models.spec.ts +++ b/packages/subagent/tool-subagent/tests/list-models.spec.ts @@ -194,7 +194,7 @@ describe('list_subagent_models', () => { ctx.llm.registerAdapter(['alpha', 'beta'], new CatalogAdapter()) expect(text(await call(ctx, {}))).toBe('alpha — ALPHA API') - expect(text(await call(ctx, { provider: 'alpha' }))).toBe('alpha/fast — Fast: Focused work.\nalpha/unlisted — Fast: Focused work.') + expect(text(await call(ctx, { provider: 'alpha' }))).toBe('alpha/fast — Fast: Focused work.') expect(text(await call(ctx, { provider: 'alpha', model: 'unlisted' }))) .toContain('alpha/unlisted — Fast') @@ -216,62 +216,92 @@ describe('list_subagent_models', () => { expect(listModels).not.toHaveBeenCalled() }) - it('resolves an allowed model even when the advertised catalog is empty', async () => { - const ctx = await setupListTool([{ provider: 'alpha', model: 'fast' }]) - ctx.llm.registerAdapter(['alpha'], new CatalogAdapter(true)) + it('returns an empty list when the catalog contains no authorized model', async () => { + const ctx = await setupListTool([{ provider: 'alpha', model: 'unlisted' }]) + ctx.llm.registerAdapter(['alpha'], new CatalogAdapter()) const result = await call(ctx, { provider: 'alpha' }) expect(result.isError).toBe(false) - expect(text(result)).toBe('alpha/fast — Fast: Focused work.') + expect(text(result)).toBe('(no advertised models for alpha)') }) - it.each([false, true])('reports each failed model alongside resolved models (all failed: %s)', async (allFailed) => { + it('lists authorized catalog entries without resolving removed or unlisted models', async () => { const ctx = await setupListTool([ { provider: 'alpha', model: 'removed' }, - ...allFailed ? [] : [{ provider: 'alpha', model: 'plain' }], - { provider: 'alpha', model: 'unavailable' }, + { provider: 'alpha', model: 'plain' }, ]) try { const adapter = new CatalogAdapter() - vi.spyOn(adapter, 'resolveModel').mockImplementation(async (provider, model) => { - if (model === 'removed') throw new Error('model was removed') - // oxlint-disable-next-line typescript/prefer-promise-reject-errors -- Exercise non-Error rejections from an external adapter. - if (model === 'unavailable') return Promise.reject('metadata unavailable') - return { provider, id: model, name: 'Plain' } - }) + const resolveModel = vi.spyOn(adapter, 'resolveModel').mockRejectedValue(new Error('lookup unavailable')) ctx.llm.registerAdapter(['alpha'], adapter) const result = await call(ctx, { provider: 'alpha' }) expect(result.isError).toBe(false) - expect(text(result)).toBe([ - 'alpha/removed — Error: model was removed', - ...allFailed ? [] : ['alpha/plain — Plain'], - 'alpha/unavailable — Error: metadata unavailable', - ].join('\n')) + expect(text(result)).toBe('alpha/plain — Plain') + expect(resolveModel).not.toHaveBeenCalled() const exact = await call(ctx, { provider: 'alpha', model: 'removed' }) expect(exact.isError).toBe(true) - expect(text(exact)).toContain('model was removed') + expect(text(exact)).toContain('lookup unavailable') } finally { await ctx.fiber.dispose() } }) - it.each([false, true])('does not render cancellation as a model error (adapter rejects: %s)', async (rejects) => { - const ctx = await setupListTool([{ provider: 'alpha', model: 'plain' }]) + it('rejects a failed catalog lookup', async () => { + const ctx = await setupListTool() + try { + const adapter = new CatalogAdapter() + vi.spyOn(adapter, 'listModels').mockRejectedValue(new Error('catalog unavailable')) + ctx.llm.registerAdapter(['alpha'], adapter) + const result = await call(ctx, { provider: 'alpha' }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('catalog unavailable') + } finally { + await ctx.fiber.dispose() + } + }) + + it('rejects cancellation after a catalog lookup', async () => { + const ctx = await setupListTool() try { const controller = new AbortController() const adapter = new CatalogAdapter() - vi.spyOn(adapter, 'resolveModel').mockImplementation(async (provider, model) => { + vi.spyOn(adapter, 'listModels').mockImplementation(async () => { controller.abort(new Error('discovery cancelled')) - if (rejects) throw new Error('lookup interrupted') - return { provider, id: model, name: 'Plain' } + return [] }) ctx.llm.registerAdapter(['alpha'], adapter) - const result = await call(ctx, { provider: 'alpha' }, controller.signal) expect(result.isError).toBe(true) expect(text(result)).toContain('discovery cancelled') - expect(text(result)).not.toContain('alpha/plain —') + } finally { + await ctx.fiber.dispose() + } + }) + + it.each([false, true])('resolves only the current parent model when selection is disabled (empty catalog: %s)', async (empty) => { + const ctx = await setup({ provider: 'mock', parentAgentOptions: { provider: 'alpha', model: 'unlisted' } }) + const adapter = new CatalogAdapter(empty) + const listModels = vi.spyOn(adapter, 'listModels') + ctx.llm.registerAdapter(['alpha'], adapter) + const result = await ctx.tools.execute({ + signal: testToolSignal, callId: ToolCallId('parent-only-discovery'), + name: 'list_subagent_models', arguments: { provider: 'alpha' }, agent: modelSelectionSetupAgent(ctx), + }) + expect(result.isError).toBe(false) + expect(text(result)).toBe('alpha/unlisted — Fast: Focused work.') + expect(listModels).not.toHaveBeenCalled() + }) + + it('requires a calling Agent when a discovery registration supplies no policy', async () => { + const ctx = new Context() + try { + await ctx.plugin(SystemPrompt) + await ctx.plugin(ToolRuntime) + registerListSubagentModels(ctx, () => undefined) + const result = await call(ctx, {}) + expect(result.isError).toBe(true) + expect(text(result)).toContain('requires a calling agent') } finally { await ctx.fiber.dispose() } diff --git a/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl b/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl index 869406b413..1a781e4739 100644 --- a/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl +++ b/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl @@ -15,7 +15,7 @@ {"type":"session/title","data":{"title":"List the models available for","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":100,"outputTokens":10},"stream":[{"type":"chunk","time":1791788384782,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"chunk","time":1791788384782,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}}},{"type":"chunk","time":1791788384782,"chunk":{"type":"usage","usage":{"inputTokens":100,"outputTokens":10}}},{"type":"chunk","time":1791788384782,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}} -{"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_discovery"},"toolCallId":"call_discovery","content":[{"type":"text","text":"discovery-fixture/removed — Error: model was removed\ndiscovery-fixture/healthy — Healthy"}],"isError":false,"id":"{{message:5}}"}},"sourceEventSeqs":[15],"surfaceOp":"append"} +{"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_discovery"},"toolCallId":"call_discovery","content":[{"type":"text","text":"discovery-fixture/healthy — Healthy"}],"isError":false,"id":"{{message:5}}"}},"sourceEventSeqs":[15],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:6}}"},"usage":{"inputTokens":100,"outputTokens":10},"stream":[{"type":"chunk","time":1791788384799,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"chunk","time":1791788384800,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1791788384800,"chunk":{"type":"usage","usage":{"inputTokens":100,"outputTokens":10}}},{"type":"chunk","time":1791788384800,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} From 155ab69fae1e2716e09f5bcb51eb30f7d61058ca Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 16:29:01 +0800 Subject: [PATCH 014/162] fix(subagent): preserve fork model selection restriction --- ...authorized-subagent-model-routes.i18n.yaml | 8 ++-- ...4-user-authorized-subagent-model-routes.md | 4 +- ...ser-authorized-subagent-model-routes.zh.md | 4 +- apps/cli/tests/web-agent-presets.e2e.ts | 4 +- docs/tool-catalog.i18n.yaml | 8 ++-- docs/tool-catalog.md | 4 +- docs/tool-catalog.zh.md | 4 +- packages/bundle/base/cordis.patch.yml | 2 +- .../subagent/tool-subagent/README.i18n.yaml | 20 ++++----- packages/subagent/tool-subagent/README.md | 10 ++--- packages/subagent/tool-subagent/README.zh.md | 10 ++--- packages/subagent/tool-subagent/src/index.ts | 10 +---- .../tests/model-selection.spec.ts | 26 ++++++++++-- .../tool-subagent/tests/tool-subagent.spec.ts | 2 +- scripts/gen-tool-catalog.ts | 2 +- .../sdk/bash-tool/tool-schemas.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../sdk/text-turn/tool-schemas.expected.json | 14 +------ .../system-prompt.expected.md | 8 +--- .../tool-schemas.expected.json | 14 +------ .../system-prompt.expected.md | 8 +--- .../tool-schemas.expected.json | 28 +------------ .../tool-schemas.expected.json | 28 +------------ .../tool-schemas.expected.json | 14 +------ .../both-mode-turn/system-prompt.expected.md | 8 +--- .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../cordis.snapshot.yml | 2 +- .../compaction-output-reserve/cordis.yml | 2 +- .../session.v4.jsonl | 2 +- .../tool-schemas.expected.json | 28 +------------ .../cordis.snapshot.yml | 2 +- .../compaction-summary-headroom/cordis.yml | 2 +- .../session.v4.jsonl | 2 +- .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 42 ++----------------- .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../lsp-definition/tool-schemas.expected.json | 14 +------ .../system-prompt.expected.md | 8 +--- .../mcp-resources/tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../system-prompt.expected.md | 8 +--- .../office-skills/tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 28 +------------ .../plugin-manager/tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../provider-cwd/tool-schemas.expected.json | 14 +------ .../system-prompt.expected.md | 8 +--- .../system-prompt.expected.md | 8 +--- .../ptc-persistent/system-prompt.expected.md | 8 +--- .../system-prompt.expected.md | 8 +--- .../ptc-python-turn/system-prompt.expected.md | 8 +--- .../ptc-read-image/system-prompt.expected.md | 8 +--- .../ptc-turn/system-prompt.expected.md | 8 +--- .../tool-schemas.expected.json | 14 +------ .../pwsh-tool-turn/tool-schemas.expected.json | 14 +------ .../ralph-loop/tool-schemas.1.expected.json | 14 +------ .../ralph-loop/tool-schemas.2.expected.json | 14 +------ .../ralph-loop/tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.1.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../text-turn/tool-schemas.expected.json | 14 +------ .../web-fetch/tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../worktree/tool-schemas.expected.json | 14 +------ .../tool-schemas.expected.json | 14 +------ .../system-prompt.expected.md | 8 +--- .../system-prompt.expected.md | 8 +--- .../web/ptc-round/system-prompt.expected.md | 8 +--- 94 files changed, 152 insertions(+), 994 deletions(-) diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index cd7f0e6dca..2592a568ad 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -9,8 +9,8 @@ en: 3238dfb7db7dfdd8 zh: ec3bdd9a9da0a321 /agent-note-user-authorized-subagent-model-routes/decision: - en: 47b020441723cc06 - zh: 8394c32b711a0b30 + en: eba9e531eb6c450b + zh: 7c0673644e83608a /agent-note-user-authorized-subagent-model-routes/alternatives-considered: en: e41c17226190ea8b zh: d3b1395fafe75f01 @@ -18,5 +18,5 @@ en: 1749cfbfda849692 zh: 9925e638cd72fba6 /agent-note-user-authorized-subagent-model-routes/related-decisions: - en: 17fdfabedb673935 - zh: 496f4f85622d7d15 + en: 0eead87ec1172d81 + zh: 4e776a583f5a00a5 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index aa0de7fad7..a91f0e9e47 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -16,7 +16,7 @@ A newly composed top-level Session snapshots the route list in `subagent/model-s The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and intersects their advertised model catalogs with the recorded route list. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts, including for ids absent from the catalog. Without a recorded policy, discovery resolves only the parent Agent’s current effective route, even if the catalog omits it. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. -Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. The primary spawn tool captures the setting once, while all capable delegation tools and shared discovery read the same Session policy. Fork defaults preserve the parent route, and its description warns about prefix recomputation when that route changes. +Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. The primary spawn tool captures the setting once, while all capable delegation tools and shared discovery read the same Session policy. Fork tools reject model-selection arguments to preserve inherited-prefix reuse. ## Alternatives considered @@ -40,4 +40,4 @@ Model selection has no unrestricted static mode. The default-off Host setting au ## Related decisions -The route arguments, adapter preflight, discovery tool, and fork cache costs remain owned by [model-selected subagent routes](../../../../packages/subagent/tool-subagent/README.md). +The route arguments, adapter preflight, discovery tool, and fork restriction remain owned by [model-selected subagent routes](../../../../packages/subagent/tool-subagent/README.md). diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index c157cec8a7..49a221d220 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -16,7 +16,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` 固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并取其公布模型目录与已记录路由列表的交集。精确查询先要求授权，再解析模型元数据及全部支持的推理强度，也允许目录未列出的 ID。没有已记录策略时，发现只解析父 Agent 当前生效的路由，即使目录未列出它。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 -模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。主 spawn 工具一次性记录设置，所有支持选择的委派工具和共享发现均读取同一 Session 策略。fork 默认保留父级路由，其描述会提示改变路由时的前缀重算成本。 +模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。主 spawn 工具一次性记录设置，所有支持选择的委派工具和共享发现均读取同一 Session 策略。fork 工具拒绝模型选择参数，以保留对继承前缀的复用。 ## Alternatives considered @@ -40,4 +40,4 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` ## Related decisions -路由参数、适配器预检、发现工具与 fork 缓存成本仍由[模型选择的 subagent 路由](../../../../packages/subagent/tool-subagent/README.zh.md)负责。 +路由参数、适配器预检、发现工具与 fork 限制仍由[模型选择的 subagent 路由](../../../../packages/subagent/tool-subagent/README.zh.md)负责。 diff --git a/apps/cli/tests/web-agent-presets.e2e.ts b/apps/cli/tests/web-agent-presets.e2e.ts index bb82f5349c..21cc43b76f 100644 --- a/apps/cli/tests/web-agent-presets.e2e.ts +++ b/apps/cli/tests/web-agent-presets.e2e.ts @@ -296,7 +296,9 @@ describe('the shipped Web composition', () => { expect(toolParameterNames(ctx, enabled.agent, 'subagent')).toEqual(expect.arrayContaining([ 'model', 'provider', 'reasoning_effort', ])) - expect(toolNames(ctx, disabled.agent)).toContain('list_subagent_models') + for (const agent of [disabled.agent, enabled.agent]) { + expect(toolParameterNames(ctx, agent, 'subagent_fork')).toEqual(['cwd', 'description', 'prompt']) + } } finally { await ctx.settings.update(SUBAGENT_MODEL_SELECTION_SETTINGS_NAMESPACE, { enabled: false }) await enabled.dispose() diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 692308776e..13a0309388 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: 6f5f5f987c218948 - zh: 018ff71df684009c + en: 559ba7c9e11d4919 + zh: 44c540c14cf43f95 /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -228,8 +228,8 @@ en: aba834ab4f70005a zh: 142f819e8da5acc7 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: c272328d36a5afaf - zh: fcf69fd0efd55ea9 + en: 0de8d49aa443ec9c + zh: 7cb7c8a55a0469b1 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 574af31642..9c94c9d1d7 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` uses that same policy, and without an enabled policy explicit selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions without inheriting parent context always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` rejects model-selection arguments; without an enabled policy, spawn selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2079,7 +2079,7 @@ Delegate a self-contained task to a subagent (a separate agent that works in its Source: [`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` uses that same policy, and without an enabled policy explicit selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. +The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions without inheriting parent context always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` rejects model-selection arguments; without an enabled policy, spawn selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index 8d36652493..be7b085212 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 使用同一策略，没有启用策略时，显式选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 且不继承父级上下文的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 拒绝模型选择参数；没有启用策略时，spawn 选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2085,7 +2085,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， 来源：[`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 使用同一策略，没有启用策略时，显式选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 +注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 且不继承父级上下文的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 拒绝模型选择参数；没有启用策略时，spawn 选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 diff --git a/packages/bundle/base/cordis.patch.yml b/packages/bundle/base/cordis.patch.yml index 9f87051c49..f162113cae 100644 --- a/packages/bundle/base/cordis.patch.yml +++ b/packages/bundle/base/cordis.patch.yml @@ -381,7 +381,7 @@ provider: spawn toolName: subagent - # Fork inherits the parent route by default for KV Cache reuse. Parent and child + # Fork omits model-selection arguments for KV Cache reuse. Parent and child # share the messaging tools; return guidance follows the inherited history. # See packages/subagent/tool-subagent/README.md # and .agents/notes/implemented/architecture/2026-08-10-fork-children-stay-one-shot.md. diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index e59a887fdd..fa0bd6c782 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -24,14 +24,14 @@ en: 42673064ba88a211 zh: 8fa26285bd649dc9 /deepseek-ai-dsh-tool-subagent/use-this-package/selecting-a-child-llm: - en: b4ba9650c228b9b6 - zh: 027833db9aafd752 + en: f85faf2e965a541a + zh: 3c232e236199b9aa /deepseek-ai-dsh-tool-subagent/understand-the-implementation: en: 6ce5f749256a7e6c zh: a4f2adb04a70d9cf /deepseek-ai-dsh-tool-subagent/understand-the-implementation/design-concept: - en: 4847e2bcdde8e282 - zh: cc833748d954a65e + en: 5c97aae02694c27a + zh: 5507f580d3e0c43b /deepseek-ai-dsh-tool-subagent/understand-the-implementation/activation-ownership: en: ef692c4d17290920 zh: b1d0d3dab011e2cd @@ -54,14 +54,14 @@ en: e4c7376d4f29dbd5 zh: 6b7a9eaf69c9ea53 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/what-the-model-sees: - en: c6da697a3965147c - zh: 4f4a48054bcbda88 + en: 56361a0d507bad66 + zh: 979d502dd3ac168f /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/token-effect: en: 6a1a6351a3a88424 zh: c574e9ee286518ed /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/kv-cache-effect: - en: 46f736c6c389052a - zh: a7765541030e9104 + en: f746a3c2f2f3cd18 + zh: 68de857073b229ff /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery: en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 @@ -114,8 +114,8 @@ en: f5d5f97a153f7c70 zh: 3d6bc05b9eadfbc2 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work: - en: 93d06d9646fe848d - zh: bc86f8bd38b4b61a + en: 897c9481685a6eab + zh: 444ee7ef4d353f60 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work/dev-note: en: e67f633dc6fe7773 zh: 3bb0c20c93f16f98 diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index 96dcfabfe6..8591e118e9 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -65,7 +65,7 @@ Every backend’s completion notice carries its final answer. Local Spawn and Fo ### Selecting a child LLM -Backends that support `agentOptions` always expose optional `provider`, `model`, and `reasoning_effort` fields. Load `@deepseek-ai/dsh-tool-subagent/list-models` once alongside the delegation tools for shared discovery; shipped compositions include it. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. Set `modelSelectionSettings: true` to sample the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. A restored Session without a policy keeps the parent-route restriction. ACP, Codex, and Claude Code do not support `agentOptions` and reject selection fields. +Backends that support `agentOptions` and do not inherit parent context always expose optional `provider`, `model`, and `reasoning_effort` fields. Load `@deepseek-ai/dsh-tool-subagent/list-models` once alongside the delegation tools for shared discovery; shipped compositions include it. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. Set `modelSelectionSettings: true` to sample the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. A restored Session without a policy keeps the parent-route restriction. ACP, Codex, and Claude Code do not support `agentOptions` and reject selection fields. A call supplies `provider` and `model` together, or supplies only an effort when configured, parent, or provider-owned defaults provide the route. Static `provider.agentRouteDefaults`, when present, form the provider/model baseline; tool configuration and model fields overlay it before route-aware effort merging and exact-route preflight. Providers without these defaults use compatible values from the parent's latest logged request, then the parent's creation options before its first request, while retaining the configured `maxTokens`. Changing the route without an explicit effort clears the inherited route-owned effort, so the selected model resolves its default. The live LLM adapter validates the effective route before child creation. Catalog membership remains advisory, so a model can use an unlisted id when its adapter accepts it. @@ -81,7 +81,7 @@ This section explains how the tool mirrors provider lifecycle and settles runs; ### Design concept -One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Settings-backed instances capture Session authorization during Agent creation; calls read the durable policy independently of registration. Direct Agent setup supplies its unpublished Session explicitly. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. Discovery has one separate composition entry shared by spawn and fork tools. No runtime invariant companion is published: policy comes directly from the Session projection and is checked at execution, with no separately retained per-Agent definition state to reconcile. +One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Settings-backed instances capture Session authorization during Agent creation; calls read the durable policy independently of registration. Direct Agent setup supplies its unpublished Session explicitly. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. Discovery has one separate composition entry. No runtime invariant companion is published: policy comes directly from the Session projection and is checked at execution, with no separately retained per-Agent definition state to reconcile. ### Activation ownership @@ -129,7 +129,7 @@ Read these pages when the package-level contract is not enough; they move from t #### What the model sees -The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. Backends supporting `agentOptions` always expose `provider`, `model`, and `reasoning_effort`. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. +The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. Backends supporting `agentOptions` that do not inherit parent context always expose `provider`, `model`, and `reasoning_effort`. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. #### Token effect @@ -137,7 +137,7 @@ Fixed schema cost per parent request, including three optional route fields on c #### KV Cache effect -Prefix-stable while provider instances and their configuration are unchanged. Adapter catalog changes do not alter the definition; a child route override may prevent a fork child from reusing the inherited parent prefix. +Prefix-stable while provider instances and their configuration are unchanged. Adapter catalog changes do not alter the definition. Fork tools omit model-selection fields to preserve reuse of the inherited parent prefix. ### Model selection and discovery @@ -209,7 +209,7 @@ Append-only; newly visible content follows the reusable request prefix and does These limits define what this tool does not return or enforce; they are current package constraints. - **This tool does not wait for results** — use the completion notice or child messages. Programmatic result waiting belongs to the subagent service. -- **Fork route changes can lose KV Cache reuse** — optional route fields follow the same Session policy as spawn; changing provider or model may require prefilling the inherited conversation again. +- **Fork tools reject model selection** — tools that inherit parent context neither expose nor accept `provider`, `model`, or `reasoning_effort`, even with Session authorization. Configured `agentOptions` remain available to the deployment. - **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection is restricted to the Session’s authorized routes, or the current parent route without an enabled policy; unsupported backends reject selection fields. diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index 0bf8ee903f..5e97c8db7a 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -65,7 +65,7 @@ kind: "package-reference" ### 选择子级 LLM -支持 `agentOptions` 的后端始终公开可选的 `provider`、`model` 和 `reasoning_effort` 字段。在委派工具旁加载一次 `@deepseek-ai/dsh-tool-subagent/list-models` 即可共享发现功能；随附组合已包含它。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。设置 `modelSelectionSettings: true` 后，新顶层 Session 会读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有策略的恢复 Session 仍受父级路由限制。ACP、Codex 和 Claude Code 不支持 `agentOptions`，会拒绝选择字段。 +支持 `agentOptions` 且不继承父级上下文的后端始终公开可选的 `provider`、`model` 和 `reasoning_effort` 字段。在委派工具旁加载一次 `@deepseek-ai/dsh-tool-subagent/list-models` 即可共享发现功能；随附组合已包含它。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。设置 `modelSelectionSettings: true` 后，新顶层 Session 会读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有策略的恢复 Session 仍受父级路由限制。ACP、Codex 和 Claude Code 不支持 `agentOptions`，会拒绝选择字段。 一次调用需同时提供 `provider` 与 `model`；当配置值、父 agent 值或提供方持有的默认值能提供路由时，也可只提供推理等级。静态的 `provider.agentRouteDefaults` 在存在时构成提供方／模型基线；工具配置与模型字段会在路由相关强度合并和确切路由预检前覆盖它。没有这些默认值的提供方会使用父 agent 最新已记录请求中的兼容值，再使用父级首次请求前的创建选项，并保留配置的 `maxTokens`。更改路由但未显式提供推理等级时，会清除继承的路由自有等级，使所选模型解析自己的默认值。实时 LLM 适配器在创建子 agent 前校验有效路由。目录成员资格只提供建议，因此适配器接受时，模型可以使用未列出的 id。 @@ -81,7 +81,7 @@ kind: "package-reference" ### 设计理念 -一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。读取设置的实例在 Agent 创建时记录 Session 授权；调用时独立于注册读取持久策略。直接配置 Agent 时显式传入尚未发布的 Session。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。发现功能由 spawn 和 fork 工具共享的独立组合项提供。本包不导出运行时不变量：策略直接来自 Session 投影，并在执行时校验，没有另行保存的逐 Agent 定义状态需要协调。 +一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。读取设置的实例在 Agent 创建时记录 Session 授权；调用时独立于注册读取持久策略。直接配置 Agent 时显式传入尚未发布的 Session。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。发现功能由一个独立组合项提供。本包不导出运行时不变量：策略直接来自 Session 投影，并在执行时校验，没有另行保存的逐 Agent 定义状态需要协调。 ### Activation 所有权 @@ -129,7 +129,7 @@ kind: "package-reference" #### 模型看到什么 -生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。支持 `agentOptions` 的后端始终公开 `provider`、`model` 和 `reasoning_effort`。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 +生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。支持 `agentOptions` 且不继承父级上下文的后端始终公开 `provider`、`model` 和 `reasoning_effort`。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 #### Token 影响 @@ -137,7 +137,7 @@ kind: "package-reference" #### KV Cache 影响 -只要提供方实例及其配置不变，前缀就保持稳定。适配器目录变化不会改变定义；子级路由覆盖可能使 fork 子 agent 无法复用继承的父级前缀。 +只要提供方实例及其配置不变，前缀就保持稳定。适配器目录变化不会改变定义。fork 工具省略模型选择字段，以保留对继承的父级前缀的复用。 ### 模型选择与发现 @@ -209,7 +209,7 @@ Start independent delegations with `subagent` or `subagent_fork` together in one 这些限制说明本工具不返回或不强制执行什么；它们是当前包约束。 - **本工具不等待结果**——使用完成通知或子任务消息。程序化结果等待由 subagent 服务提供。 -- **更改 fork 路由可能失去 KV Cache 复用** — 可选路由字段遵循与 spawn 相同的 Session 策略；更改提供方或模型可能需要重新预填充继承的对话。 +- **fork 工具拒绝模型选择** — 继承父级上下文的工具既不公开也不接受 `provider`、`model` 或 `reasoning_effort`，即使 Session 已授权也不例外。部署仍可配置 `agentOptions`。 - **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择受 Session 授权路由限制，没有启用策略时只允许当前父级路由；不支持选择的后端会拒绝选择字段。 diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index 90f98d10aa..6cba7d3490 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -206,18 +206,12 @@ export function apply(ctx: Context, config: Config, session?: Session): void { let mounted: { subagentProvider: SubagentProvider; disposeTool: () => void } | undefined const mount = (subagentProvider: SubagentProvider): void => { assertSubagentProviderConfiguration(subagentProvider) - const modelSelectionEnabled = subagentProvider.capabilities.agentOptions + const modelSelectionEnabled = subagentProvider.capabilities.agentOptions && !subagentProvider.inheritsParentContext const wording = providerWording(subagentProvider.inheritsParentContext) const providerRouteDefaults = subagentProvider.agentRouteDefaults const selectionDescription = providerRouteDefaults !== undefined ? ' Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider\'s route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model\'s default effort.' : ' Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model\'s default effort.' - const choiceDescription = !modelSelectionEnabled - ? '' - : selectionDescription - + (subagentProvider.inheritsParentContext - ? ' Changing the route can prevent provider-side reuse of the inherited conversation prefix.' - : '') const definition = defineTool({ name: toolName, description: wording.description @@ -225,7 +219,7 @@ export function apply(ctx: Context, config: Config, session?: Session): void { + (subagentProvider.prepareContinuable !== undefined ? ' The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.' : ' The completion notice includes its final answer. This backend does not accept follow-up messages.') - + choiceDescription, + + (modelSelectionEnabled ? selectionDescription : ''), parameters: { cwd: { type: 'string', diff --git a/packages/subagent/tool-subagent/tests/model-selection.spec.ts b/packages/subagent/tool-subagent/tests/model-selection.spec.ts index 4af0606e66..4a99144c81 100644 --- a/packages/subagent/tool-subagent/tests/model-selection.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection.spec.ts @@ -469,11 +469,31 @@ describe('dsh-tool-subagent model selection', () => { expect(resolve).not.toHaveBeenCalled() }) - it('warns that changing a fork route can lose inherited-prefix reuse', async () => { - const ctx = await setup({ provider: 'mock', withModelSelection: true }, { inheritsParentContext: true }) + it('keeps fork model selection unavailable even with Session authorization', async () => { + const started = vi.fn() + const ctx = await setup({ provider: 'mock', withModelSelection: true }, { + inheritsParentContext: true, + onStart: started, + }) const schema = ctx.tools.schemas(modelSelectionSetupAgent(ctx)).find(entry => entry.name === 'subagent')! + const props = (schema.parameters as { properties?: Record }).properties ?? {} + expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'prompt']) expect(schema.description).toContain('inherits this conversation') - expect(schema.description).toContain('can prevent provider-side reuse of the inherited conversation prefix') + expect(schema.description).not.toContain('Child LLM selection') + + for (const selection of [ + { provider: 'alpha', model: 'fast-model' }, + { reasoning_effort: 'high' }, + ]) { + const result = await callSubagent(ctx, { description: 'fork work', prompt: 'do it', ...selection }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('model selection is disabled for this tool instance') + } + expect(started).not.toHaveBeenCalled() + + const inherited = await callSubagent(ctx, { description: 'fork work', prompt: 'do it' }) + expect(inherited.isError).toBe(false) + expect(started).toHaveBeenCalledOnce() }) it('propagates an exact-route resolver failure before child creation', async () => { diff --git a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts index 569b1dcb7b..9ddecc530f 100644 --- a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts +++ b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts @@ -513,7 +513,7 @@ describe('dsh-tool-subagent', () => { const schema = ctx.tools.schemas().find(s => s.name === 'subagent')! expect(schema.description).toContain('inherits this conversation') expect(schema.description).not.toContain('does not see this conversation') - expect(schema.description).toContain('can prevent provider-side reuse of the inherited conversation prefix') + expect(schema.description).not.toContain('Child LLM selection') const props = (schema.parameters as { properties: Record }).properties expect(props['prompt']!.description).toContain('completed turns') }) diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index fa03619cba..81b10491e0 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -522,7 +522,7 @@ const TOOL_PACKAGES: ToolPackage[] = [ await ctx.plugin(SubagentModelDiscovery) }, note: - 'The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` uses that same policy, and without an enabled policy explicit selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', + 'The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions without inheriting parent context always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` rejects model-selection arguments; without an enabled policy, spawn selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', }, { pkg: '@deepseek-ai/dsh-tool-subagent-control', diff --git a/snapshots/sdk/bash-tool/tool-schemas.expected.json b/snapshots/sdk/bash-tool/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/bash-tool/tool-schemas.expected.json +++ b/snapshots/sdk/bash-tool/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json index 17dff0816f..e13cfd3bb0 100644 --- a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json @@ -467,7 +467,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -482,18 +482,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json index 87515e6d78..5f0e685eb2 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/text-turn/tool-schemas.expected.json b/snapshots/sdk/text-turn/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/sdk/text-turn/tool-schemas.expected.json +++ b/snapshots/sdk/text-turn/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/sdk/tool-error-details/system-prompt.expected.md b/snapshots/sdk/tool-error-details/system-prompt.expected.md index 08bc866504..4dcf8262a6 100644 --- a/snapshots/sdk/tool-error-details/system-prompt.expected.md +++ b/snapshots/sdk/tool-error-details/system-prompt.expected.md @@ -221,7 +221,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -229,12 +229,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/sdk/tool-error-details/tool-schemas.expected.json b/snapshots/sdk/tool-error-details/tool-schemas.expected.json index 62d25abd75..77b5236350 100644 --- a/snapshots/sdk/tool-error-details/tool-schemas.expected.json +++ b/snapshots/sdk/tool-error-details/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/advanced-toolchain/system-prompt.expected.md b/snapshots/session/advanced-toolchain/system-prompt.expected.md index d3a212e8ff..14d8e2559d 100644 --- a/snapshots/session/advanced-toolchain/system-prompt.expected.md +++ b/snapshots/session/advanced-toolchain/system-prompt.expected.md @@ -234,7 +234,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -242,12 +242,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/session/advanced-toolchain/tool-schemas.expected.json b/snapshots/session/advanced-toolchain/tool-schemas.expected.json index 46396bb306..f8cbf6b810 100644 --- a/snapshots/session/advanced-toolchain/tool-schemas.expected.json +++ b/snapshots/session/advanced-toolchain/tool-schemas.expected.json @@ -434,7 +434,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -449,18 +449,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1188,7 +1176,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1203,18 +1191,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/agent-instructions/tool-schemas.expected.json b/snapshots/session/agent-instructions/tool-schemas.expected.json index 472458e753..5018eac594 100644 --- a/snapshots/session/agent-instructions/tool-schemas.expected.json +++ b/snapshots/session/agent-instructions/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1067,7 +1055,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1082,18 +1070,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/background-confinement-failure/tool-schemas.expected.json b/snapshots/session/background-confinement-failure/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/background-confinement-failure/tool-schemas.expected.json +++ b/snapshots/session/background-confinement-failure/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/both-mode-turn/system-prompt.expected.md b/snapshots/session/both-mode-turn/system-prompt.expected.md index 5e49146818..69bc221437 100644 --- a/snapshots/session/both-mode-turn/system-prompt.expected.md +++ b/snapshots/session/both-mode-turn/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json index a8ef352d5e..763c4962d9 100644 --- a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json @@ -500,7 +500,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -515,18 +515,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json index 3d46faeef4..f8db659ff3 100644 --- a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json @@ -502,7 +502,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -517,18 +517,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json index 33c0bdbb5a..f3d955b641 100644 --- a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json +++ b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json @@ -606,7 +606,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -621,18 +621,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/claude-code-mods/tool-schemas.expected.json b/snapshots/session/claude-code-mods/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/claude-code-mods/tool-schemas.expected.json +++ b/snapshots/session/claude-code-mods/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/compaction-output-reserve/cordis.snapshot.yml b/snapshots/session/compaction-output-reserve/cordis.snapshot.yml index 15f69e1913..b9294ef273 100644 --- a/snapshots/session/compaction-output-reserve/cordis.snapshot.yml +++ b/snapshots/session/compaction-output-reserve/cordis.snapshot.yml @@ -31,7 +31,7 @@ name: DeepSeek models: - id: deepseek-v4-flash - contextWindow: 9270 + contextWindow: 9009 defaultMaxTokens: 1000 - id: compaction-working-directory name: '../../../packages/test-support/session-snapshot/tests/fixtures/compaction-working-directory.ts' diff --git a/snapshots/session/compaction-output-reserve/cordis.yml b/snapshots/session/compaction-output-reserve/cordis.yml index ea561fc74f..e970dde73c 100644 --- a/snapshots/session/compaction-output-reserve/cordis.yml +++ b/snapshots/session/compaction-output-reserve/cordis.yml @@ -7,7 +7,7 @@ config: models: - id: deepseek-v4-flash - contextWindow: 9270 + contextWindow: 9009 maxTokens: 1000 - id: compaction-basic diff --git a/snapshots/session/compaction-output-reserve/session.v4.jsonl b/snapshots/session/compaction-output-reserve/session.v4.jsonl index 590b3b6ac9..51414cb663 100644 --- a/snapshots/session/compaction-output-reserve/session.v4.jsonl +++ b/snapshots/session/compaction-output-reserve/session.v4.jsonl @@ -11,7 +11,7 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Establish a durable compaction premise before continuing. Record every part of this historical evidence: the snapshot uses keyless replay; persistence uses JSONL; the assembled headless application loads its real Cordis composition; model-visible inputs remain logged; tool calls and results remain paired and ordered; output reservation reduces the message budget before proactive compaction; compaction opens with compaction/start and closes with compaction/end; a successful auxiliary summary records compaction/summary source metadata; the replacement surface shadows only an older balanced range; the checkpoint remains smaller than the history it replaces; the newest tool result remains verbatim; the next request sees that checkpoint; the final response proves the same turn continued; deterministic snapshot evidence stays separate from the live-provider smoke; and no external API key is needed. Emit one alpha marker through bash, then finish the task after proactive compaction with the exact words PROACTIVE COMPACTION COMPLETE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"/tmp\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"/tmp\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":1000},"adapterDefaults":{"maxTokens":true},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9270}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9009}} {"type":"session/title","data":{"title":"Establish a durable compaction premise","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":24,"outputTokens":6},"stream":[{"type":"chunk","time":1791427805327,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791427805327,"index":0,"dt":[],"id":"call_compaction_marker","name":"bash","args":["{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"]},{"type":"chunk","time":1791427805327,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}}},{"type":"chunk","time":1791427805327,"chunk":{"type":"usage","usage":{"inputTokens":24,"outputTokens":6}}},{"type":"chunk","time":1791427805327,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}} diff --git a/snapshots/session/compaction-recovery/tool-schemas.expected.json b/snapshots/session/compaction-recovery/tool-schemas.expected.json index 472458e753..5018eac594 100644 --- a/snapshots/session/compaction-recovery/tool-schemas.expected.json +++ b/snapshots/session/compaction-recovery/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1067,7 +1055,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1082,18 +1070,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml b/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml index 76355f3bb4..342f2c3574 100644 --- a/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml +++ b/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml @@ -30,7 +30,7 @@ name: DeepSeek models: - id: deepseek-v4-flash - contextWindow: 9270 + contextWindow: 9009 defaultMaxTokens: 1000 - id: compaction-working-directory name: '../../../packages/test-support/session-snapshot/tests/fixtures/compaction-working-directory.ts' diff --git a/snapshots/session/compaction-summary-headroom/cordis.yml b/snapshots/session/compaction-summary-headroom/cordis.yml index da55c9afc7..1897460f05 100644 --- a/snapshots/session/compaction-summary-headroom/cordis.yml +++ b/snapshots/session/compaction-summary-headroom/cordis.yml @@ -7,7 +7,7 @@ config: models: - id: deepseek-v4-flash - contextWindow: 9270 + contextWindow: 9009 maxTokens: 1000 - id: compaction-basic diff --git a/snapshots/session/compaction-summary-headroom/session.v4.jsonl b/snapshots/session/compaction-summary-headroom/session.v4.jsonl index 3a4e672909..aafe00a914 100644 --- a/snapshots/session/compaction-summary-headroom/session.v4.jsonl +++ b/snapshots/session/compaction-summary-headroom/session.v4.jsonl @@ -11,7 +11,7 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Establish a durable compaction premise before continuing. Record every part of this historical evidence: the snapshot uses keyless replay; persistence uses JSONL; the assembled headless application loads its real Cordis composition; model-visible inputs remain logged; tool calls and results remain paired and ordered; output reservation reduces the message budget before proactive compaction; compaction opens with compaction/start and closes with compaction/end; a successful auxiliary summary records compaction/summary source metadata; the replacement surface shadows only an older balanced range; the checkpoint remains smaller than the history it replaces; the newest tool result remains verbatim; the next request sees that checkpoint; the final response proves the same turn continued; deterministic snapshot evidence stays separate from the live-provider smoke; and no external API key is needed. Emit one alpha marker through bash, then finish the task after proactive compaction with the exact words PROACTIVE COMPACTION COMPLETE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"/tmp\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"/tmp\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":1000},"adapterDefaults":{"maxTokens":true},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9270}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9009}} {"type":"session/title","data":{"title":"Establish a durable compaction premise","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":24,"outputTokens":6},"stream":[{"type":"chunk","time":1791427807038,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791427807038,"index":0,"dt":[],"id":"call_compaction_marker","name":"bash","args":["{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"]},{"type":"chunk","time":1791427807038,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}}},{"type":"chunk","time":1791427807038,"chunk":{"type":"usage","usage":{"inputTokens":24,"outputTokens":6}}},{"type":"chunk","time":1791427807038,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}} diff --git a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json index 9a7dc19de7..1107b15b8f 100644 --- a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json @@ -479,7 +479,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -494,18 +494,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json index a95daa24a6..0bf5c52da2 100644 --- a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json @@ -439,7 +439,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -454,18 +454,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json index da58565d4e..e8770b87d3 100644 --- a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json +++ b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json @@ -434,7 +434,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -449,18 +449,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json index 5268b0e9d8..de74610301 100644 --- a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json +++ b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1075,7 +1063,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1090,18 +1078,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1748,7 +1724,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1763,18 +1739,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json +++ b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json index 840733798a..3fb11acd59 100644 --- a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json +++ b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json @@ -297,7 +297,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -312,18 +312,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/lsp-definition/tool-schemas.expected.json b/snapshots/session/lsp-definition/tool-schemas.expected.json index 473b6cc111..c9da49006c 100644 --- a/snapshots/session/lsp-definition/tool-schemas.expected.json +++ b/snapshots/session/lsp-definition/tool-schemas.expected.json @@ -430,7 +430,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -445,18 +445,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md index 48f67a9d80..da3f0b8ab7 100644 --- a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md +++ b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md @@ -217,7 +217,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -225,12 +225,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/session/mcp-resources/tool-schemas.expected.json b/snapshots/session/mcp-resources/tool-schemas.expected.json index 5a7f33f534..b7e47f91c6 100644 --- a/snapshots/session/mcp-resources/tool-schemas.expected.json +++ b/snapshots/session/mcp-resources/tool-schemas.expected.json @@ -454,7 +454,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -469,18 +469,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json index dec41c176b..8cfd4f8c9f 100644 --- a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json +++ b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json @@ -412,7 +412,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -427,18 +427,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/office-progressive-native/tool-schemas.expected.json b/snapshots/session/office-progressive-native/tool-schemas.expected.json index 7269632ffd..f43f78ab5f 100644 --- a/snapshots/session/office-progressive-native/tool-schemas.expected.json +++ b/snapshots/session/office-progressive-native/tool-schemas.expected.json @@ -348,7 +348,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -363,18 +363,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/office-progressive-ptc/system-prompt.expected.md b/snapshots/session/office-progressive-ptc/system-prompt.expected.md index e56a80173c..c5fc982e04 100644 --- a/snapshots/session/office-progressive-ptc/system-prompt.expected.md +++ b/snapshots/session/office-progressive-ptc/system-prompt.expected.md @@ -167,7 +167,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -175,12 +175,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/session/office-skills/tool-schemas.expected.json b/snapshots/session/office-skills/tool-schemas.expected.json index 7269632ffd..f43f78ab5f 100644 --- a/snapshots/session/office-skills/tool-schemas.expected.json +++ b/snapshots/session/office-skills/tool-schemas.expected.json @@ -348,7 +348,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -363,18 +363,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json index bee9fc050c..eaefa6b809 100644 --- a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json @@ -268,7 +268,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -283,18 +283,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json index bc73a4b1fb..c06f67618d 100644 --- a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json @@ -454,7 +454,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -469,18 +469,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ @@ -1259,7 +1247,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1274,18 +1262,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/plugin-manager/tool-schemas.expected.json b/snapshots/session/plugin-manager/tool-schemas.expected.json index 47b4c83859..6357a8beab 100644 --- a/snapshots/session/plugin-manager/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager/tool-schemas.expected.json @@ -454,7 +454,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -469,18 +469,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/product-subagent-both/tool-schemas.expected.json b/snapshots/session/product-subagent-both/tool-schemas.expected.json index 779cf01085..ad221ee080 100644 --- a/snapshots/session/product-subagent-both/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-both/tool-schemas.expected.json @@ -493,7 +493,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -508,18 +508,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/product-subagent-codex/tool-schemas.expected.json b/snapshots/session/product-subagent-codex/tool-schemas.expected.json index 40da027b47..7df665c825 100644 --- a/snapshots/session/product-subagent-codex/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-codex/tool-schemas.expected.json @@ -443,7 +443,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -458,18 +458,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json index d202776234..50fbaad047 100644 --- a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json @@ -418,7 +418,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -433,18 +433,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/provider-cwd/tool-schemas.expected.json b/snapshots/session/provider-cwd/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/provider-cwd/tool-schemas.expected.json +++ b/snapshots/session/provider-cwd/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ptc-node-read-only/system-prompt.expected.md b/snapshots/session/ptc-node-read-only/system-prompt.expected.md index 835372b136..43224a84f3 100644 --- a/snapshots/session/ptc-node-read-only/system-prompt.expected.md +++ b/snapshots/session/ptc-node-read-only/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { diff --git a/snapshots/session/ptc-node-workspace/system-prompt.expected.md b/snapshots/session/ptc-node-workspace/system-prompt.expected.md index 835372b136..43224a84f3 100644 --- a/snapshots/session/ptc-node-workspace/system-prompt.expected.md +++ b/snapshots/session/ptc-node-workspace/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { diff --git a/snapshots/session/ptc-persistent/system-prompt.expected.md b/snapshots/session/ptc-persistent/system-prompt.expected.md index 835372b136..43224a84f3 100644 --- a/snapshots/session/ptc-persistent/system-prompt.expected.md +++ b/snapshots/session/ptc-persistent/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { diff --git a/snapshots/session/ptc-policy-change/system-prompt.expected.md b/snapshots/session/ptc-policy-change/system-prompt.expected.md index 8308b029e0..07099f54b6 100644 --- a/snapshots/session/ptc-policy-change/system-prompt.expected.md +++ b/snapshots/session/ptc-policy-change/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/session/ptc-python-turn/system-prompt.expected.md b/snapshots/session/ptc-python-turn/system-prompt.expected.md index 41a8a14056..45143cc82e 100644 --- a/snapshots/session/ptc-python-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-python-turn/system-prompt.expected.md @@ -396,12 +396,6 @@ class SubagentForkArgs(TypedDict): description: str # The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. prompt: str - # LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. - provider: NotRequired[str] - # Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. - model: NotRequired[str] - # Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. - reasoning_effort: NotRequired[str] # Additional keys beyond those declared are allowed. class SubagentForkOutput(TypedDict): @@ -567,7 +561,7 @@ class Tools(Protocol): async def subagent(self, args: SubagentArgs) -> SubagentOutput: """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.""" async def subagent_fork(self, args: SubagentForkArgs) -> SubagentForkOutput: - """Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.""" + """Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.""" async def todo_write(self, args: TodoWriteArgs) -> TodoWriteOutput: """Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done.""" async def update_goal(self, args: UpdateGoalArgs) -> UpdateGoalOutput1 | UpdateGoalOutput2: diff --git a/snapshots/session/ptc-read-image/system-prompt.expected.md b/snapshots/session/ptc-read-image/system-prompt.expected.md index 38a2fa9fbc..0af031b0c8 100644 --- a/snapshots/session/ptc-read-image/system-prompt.expected.md +++ b/snapshots/session/ptc-read-image/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/session/ptc-turn/system-prompt.expected.md b/snapshots/session/ptc-turn/system-prompt.expected.md index 835372b136..43224a84f3 100644 --- a/snapshots/session/ptc-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-turn/system-prompt.expected.md @@ -188,7 +188,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -196,12 +196,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Echo template text such as {{item}}, {{model}}, or {{ item }} without substitution. */ template_echo: { diff --git a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json index d33cb92a78..816b7fcb6a 100644 --- a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json +++ b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json index dcf23f6b31..f9cd850b97 100644 --- a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json @@ -349,7 +349,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -364,18 +364,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.1.expected.json b/snapshots/session/ralph-loop/tool-schemas.1.expected.json index 93d4b861a1..0bb4c382a8 100644 --- a/snapshots/session/ralph-loop/tool-schemas.1.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.1.expected.json @@ -456,7 +456,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -471,18 +471,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.2.expected.json b/snapshots/session/ralph-loop/tool-schemas.2.expected.json index 93d4b861a1..0bb4c382a8 100644 --- a/snapshots/session/ralph-loop/tool-schemas.2.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.2.expected.json @@ -456,7 +456,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -471,18 +471,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.expected.json b/snapshots/session/ralph-loop/tool-schemas.expected.json index 44c18da093..f67a9cc249 100644 --- a/snapshots/session/ralph-loop/tool-schemas.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.expected.json @@ -413,7 +413,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -428,18 +428,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/session-query-spill/tool-schemas.expected.json b/snapshots/session/session-query-spill/tool-schemas.expected.json index dd5544d7d4..18b7f4a24d 100644 --- a/snapshots/session/session-query-spill/tool-schemas.expected.json +++ b/snapshots/session/session-query-spill/tool-schemas.expected.json @@ -597,7 +597,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -612,18 +612,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json index f61fc21202..233a8b87a8 100644 --- a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json @@ -418,7 +418,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -433,18 +433,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json index cd84466ebf..ac4d94f100 100644 --- a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json +++ b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json @@ -456,7 +456,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -471,18 +471,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json index 960e60afd0..045312a524 100644 --- a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json @@ -286,7 +286,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -301,18 +301,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json index 3e450f980e..9c47c0c649 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json @@ -408,7 +408,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -423,18 +423,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json index 8ebad49ed3..2eaec5d921 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json index c38ea849b9..f9cbf2b0d8 100644 --- a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json @@ -307,7 +307,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -322,18 +322,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json +++ b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/text-turn/tool-schemas.expected.json b/snapshots/session/text-turn/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/text-turn/tool-schemas.expected.json +++ b/snapshots/session/text-turn/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/web-fetch/tool-schemas.expected.json b/snapshots/session/web-fetch/tool-schemas.expected.json index 11dba4b997..40cc9548ef 100644 --- a/snapshots/session/web-fetch/tool-schemas.expected.json +++ b/snapshots/session/web-fetch/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/workflow-confinement/tool-schemas.expected.json b/snapshots/session/workflow-confinement/tool-schemas.expected.json index 2dd1f341df..8dc90581f4 100644 --- a/snapshots/session/workflow-confinement/tool-schemas.expected.json +++ b/snapshots/session/workflow-confinement/tool-schemas.expected.json @@ -393,7 +393,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +408,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/workspace-dependencies/tool-schemas.expected.json b/snapshots/session/workspace-dependencies/tool-schemas.expected.json index cca2b4d36a..9bd9bd41b5 100644 --- a/snapshots/session/workspace-dependencies/tool-schemas.expected.json +++ b/snapshots/session/workspace-dependencies/tool-schemas.expected.json @@ -356,7 +356,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +371,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/session/worktree/tool-schemas.expected.json b/snapshots/session/worktree/tool-schemas.expected.json index a18f81001c..dd66d0a51f 100644 --- a/snapshots/session/worktree/tool-schemas.expected.json +++ b/snapshots/session/worktree/tool-schemas.expected.json @@ -410,7 +410,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -425,18 +425,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/web/fresh-round-trip/tool-schemas.expected.json b/snapshots/web/fresh-round-trip/tool-schemas.expected.json index 02a74c4f52..26471f7526 100644 --- a/snapshots/web/fresh-round-trip/tool-schemas.expected.json +++ b/snapshots/web/fresh-round-trip/tool-schemas.expected.json @@ -756,7 +756,7 @@ }, { "name": "subagent_fork", - "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix.", + "description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -771,18 +771,6 @@ "prompt": { "type": "string", "description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new." - }, - "provider": { - "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." - }, - "model": { - "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." - }, - "reasoning_effort": { - "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." } }, "required": [ diff --git a/snapshots/web/official-subagent-bundle/system-prompt.expected.md b/snapshots/web/official-subagent-bundle/system-prompt.expected.md index 8908ece5a4..dd702b72a4 100644 --- a/snapshots/web/official-subagent-bundle/system-prompt.expected.md +++ b/snapshots/web/official-subagent-bundle/system-prompt.expected.md @@ -312,7 +312,7 @@ interface ToolArgsMap { /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -320,12 +320,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/web/optional-string-editor/system-prompt.expected.md b/snapshots/web/optional-string-editor/system-prompt.expected.md index 00fa427842..597ebe90cd 100644 --- a/snapshots/web/optional-string-editor/system-prompt.expected.md +++ b/snapshots/web/optional-string-editor/system-prompt.expected.md @@ -320,7 +320,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -328,12 +328,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { diff --git a/snapshots/web/ptc-round/system-prompt.expected.md b/snapshots/web/ptc-round/system-prompt.expected.md index 0c53fb848d..731d7cc7b4 100644 --- a/snapshots/web/ptc-round/system-prompt.expected.md +++ b/snapshots/web/ptc-round/system-prompt.expected.md @@ -303,7 +303,7 @@ interface ToolArgsMap { /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ reasoning_effort?: string; } & Record; - /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. Changing the route can prevent provider-side reuse of the inherited conversation prefix. */ + /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -311,12 +311,6 @@ interface ToolArgsMap { description: string; /** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ - provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ - model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ - reasoning_effort?: string; } & Record; /** Record and update a task list to plan multi-step work and show progress; skip it for trivial single-step tasks. Add one todo per concrete step before you start. While work remains, keep the todos being worked on `in_progress`, several only when work runs in parallel. Mark each todo `completed` as soon as it is done. */ todo_write: { From 56affefb19d9f6888cd90f22295a3bb3b2801acf Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 17:20:06 +0800 Subject: [PATCH 015/162] refactor(subagent): shorten model-selection tool descriptions --- docs/tool-catalog.i18n.yaml | 8 ++-- docs/tool-catalog.md | 14 +++---- docs/tool-catalog.zh.md | 14 +++---- packages/subagent/tool-subagent/src/index.ts | 16 +++---- .../subagent/tool-subagent/src/list-models.ts | 12 +++--- .../tool-subagent/tests/tool-subagent.spec.ts | 4 +- .../sdk/bash-tool/tool-schemas.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.expected.json | 30 ++++++------- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../sdk/text-turn/tool-schemas.expected.json | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../tool-schemas.expected.json | 28 ++++++------- .../tool-schemas.expected.json | 28 ++++++------- .../tool-schemas.expected.json | 14 +++---- .../both-mode-turn/system-prompt.expected.md | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 28 ++++++------- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 42 +++++++++---------- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../lsp-definition/tool-schemas.expected.json | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../mcp-resources/tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../office-skills/tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 28 ++++++------- .../plugin-manager/tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../provider-cwd/tool-schemas.expected.json | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../ptc-persistent/system-prompt.expected.md | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../ptc-python-turn/system-prompt.expected.md | 14 +++---- .../ptc-read-image/system-prompt.expected.md | 14 +++---- .../ptc-turn/system-prompt.expected.md | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../pwsh-tool-turn/tool-schemas.expected.json | 14 +++---- .../ralph-loop/tool-schemas.1.expected.json | 14 +++---- .../ralph-loop/tool-schemas.2.expected.json | 14 +++---- .../ralph-loop/tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.1.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../team-targets/tool-schemas.expected.json | 6 +-- .../text-turn/tool-schemas.expected.json | 14 +++---- .../web-fetch/tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../worktree/tool-schemas.expected.json | 14 +++---- .../tool-schemas.expected.json | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../system-prompt.expected.md | 14 +++---- .../web/ptc-round/system-prompt.expected.md | 14 +++---- 80 files changed, 594 insertions(+), 602 deletions(-) diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 13a0309388..4b23c19b5c 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -225,11 +225,11 @@ en: 841abe0dd5913ea7 zh: 841abe0dd5913ea7 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/list-subagent-models: - en: aba834ab4f70005a - zh: 142f819e8da5acc7 + en: 6872bee11fccb6c8 + zh: 230f2d443dc5d9a6 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: 0de8d49aa443ec9c - zh: 7cb7c8a55a0469b1 + en: aff4a6668b41b7d5 + zh: 98d6ddf8aa28c7f9 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 9c94c9d1d7..769ad76242 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -2017,7 +2017,7 @@ The five read-only tools hide provider cursors and authorize every result from t ### `list_subagent_models` -Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. +Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. ```json { @@ -2025,11 +2025,11 @@ Discover LLM routes for subagents without changing the current Agent. Call with "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -2039,7 +2039,7 @@ Source: [`packages/subagent/tool-subagent/src/list-models.ts`](../packages/subag ### `subagent` -Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. +Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. ```json { @@ -2059,15 +2059,15 @@ Delegate a self-contained task to a subagent (a separate agent that works in its }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index be7b085212..a91c05a5cb 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -2023,7 +2023,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， ### `list_subagent_models` -发现 subagent 可用的 LLM 路由，不更改当前 Agent。无参数调用会列出已注册提供方；提供 `provider` 时会列出其允许的模型；同时提供 `provider` 和 `model` 时会检查该精确模型及其推理强度。目录条目只提供建议：adapter 可能接受未列出的模型 id。把返回的 id 用于委派工具的 `provider`、`model` 与 `reasoning_effort` 字段。 +查询 subagent 可用的模型，不更改当前 Agent。无参数：提供方；`provider`：允许的模型；`provider` + `model`：该模型及其推理强度。将返回的 id 用于委派。 ```json { @@ -2031,11 +2031,11 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -2045,7 +2045,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， ### `subagent` -将一项自包含任务委派给 subagent（在自身上下文中工作的独立 agent），用它卸载聚焦且独立的工作，例如研究、限定范围的实现或分析，以免消耗当前对话的上下文。subagent 会返回结果，但不会返回中间步骤。此工具启动独立管理的 subagent 并立即返回其 id。运行时会在任务结束时通知你。子代理通过 `send_message` 汇报结果；你可以使用 `send_message` 在运行时引导它，或在结束后继续对话。子级 LLM 选择是可选的。省略 `provider`、`model` 和 `reasoning_effort` 会使用配置的子级默认值，并从父 Agent 继承缺失且兼容的值。先通过 `list_subagent_models` 查询路由和强度，再同时提供 `provider` 与 `model`。更改生效路由却不指定强度时，使用所选模型的默认强度。 +将一项自包含任务委派给 subagent（在自身上下文中工作的独立 agent），用它卸载聚焦且独立的工作，例如研究、限定范围的实现或分析，以免消耗当前对话的上下文。subagent 会返回结果，但不会返回中间步骤。此工具启动独立管理的 subagent 并立即返回其 id。运行时会在任务结束时通知你。子代理通过 `send_message` 汇报结果；你可以使用 `send_message` 在运行时引导它，或在结束后继续对话。先通过 `list_subagent_models` 查询选项，再同时设置 `provider` 与 `model`。省略的值使用子级默认值，缺失时使用兼容的父级值。生效路由改变时，省略 `reasoning_effort` 会使用模型的默认强度。 ```json { @@ -2065,15 +2065,15 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index 6cba7d3490..19af35e38f 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -210,8 +210,8 @@ export function apply(ctx: Context, config: Config, session?: Session): void { const wording = providerWording(subagentProvider.inheritsParentContext) const providerRouteDefaults = subagentProvider.agentRouteDefaults const selectionDescription = providerRouteDefaults !== undefined - ? ' Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider\'s route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model\'s default effort.' - : ' Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model\'s default effort.' + ? ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend\'s route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' + : ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' const definition = defineTool({ name: toolName, description: wording.description @@ -238,21 +238,15 @@ export function apply(ctx: Context, config: Config, session?: Session): void { ...modelSelectionEnabled ? { provider: { type: 'string' as const, - description: providerRouteDefaults !== undefined - ? 'LLM provider route for the child. Supply together with model; omit both to use configured child defaults or this provider\'s route defaults.' - : 'LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route.', + description: 'Child LLM provider id.', }, model: { type: 'string' as const, - description: providerRouteDefaults !== undefined - ? 'Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or this provider\'s route defaults.' - : 'Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route.', + description: 'Model id within that provider.', }, reasoning_effort: { type: 'string' as const, - description: providerRouteDefaults !== undefined - ? 'Adapter-owned reasoning effort for the effective child route. Omit to use a compatible configured effort or the selected model\'s default.' - : 'Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model\'s default.', + description: 'Model-specific reasoning effort id.', }, } : {}, diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index 950cf09e72..5c100eedf6 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -104,19 +104,17 @@ export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolR ctx.tools.register(defineTool({ name: 'list_subagent_models', description: - 'Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list ' - + 'registered providers, with `provider` to list its allowed models, or with `provider` and `model` ' - + 'to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may ' - + 'accept an unlisted model id. Use the returned ids with a delegation tool\'s `provider`, `model`, and ' - + '`reasoning_effort` fields.', + 'Inspect subagent models without changing the current Agent. No arguments: providers; ' + + '`provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. ' + + 'Use returned ids for delegation.', parameters: { provider: { type: 'string', - description: 'Registered LLM provider id. Omit to list providers.', + description: 'Registered LLM provider id.', }, model: { type: 'string', - description: 'Exact model id to inspect. Requires provider; omit to list that provider\'s allowed models.', + description: 'Exact model id; requires provider.', }, }, output: { diff --git a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts index 9ddecc530f..4a0a1a6a08 100644 --- a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts +++ b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts @@ -266,7 +266,7 @@ describe('dsh-tool-subagent', () => { model: 'child-model', }) expect(ctx.tools.schemas(modelSelectionSetupAgent(ctx)).find(schema => schema.name === 'subagent')?.description) - .toContain('this provider\'s route defaults') + .toContain('this backend\'s route defaults') expect(seen?.agentOptions).toEqual({ provider: 'alpha', model: 'child-model', @@ -513,7 +513,7 @@ describe('dsh-tool-subagent', () => { const schema = ctx.tools.schemas().find(s => s.name === 'subagent')! expect(schema.description).toContain('inherits this conversation') expect(schema.description).not.toContain('does not see this conversation') - expect(schema.description).not.toContain('Child LLM selection') + expect(schema.description).not.toContain('Inspect options with `list_subagent_models`') const props = (schema.parameters as { properties: Record }).properties expect(props['prompt']!.description).toContain('completed turns') }) diff --git a/snapshots/sdk/bash-tool/tool-schemas.expected.json b/snapshots/sdk/bash-tool/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/bash-tool/tool-schemas.expected.json +++ b/snapshots/sdk/bash-tool/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json index e13cfd3bb0..ef3f413d82 100644 --- a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -393,7 +393,7 @@ }, { "name": "subagent_dsh_sdk", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider's route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -411,15 +411,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or this provider's route defaults." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or this provider's route defaults." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to use a compatible configured effort or the selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -430,7 +430,7 @@ }, { "name": "subagent_dsh_sdk_second", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider's route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -448,15 +448,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or this provider's route defaults." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or this provider's route defaults." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to use a compatible configured effort or the selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json index 5f0e685eb2..85e10ae4f2 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and this provider's route defaults. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or this provider's route defaults." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or this provider's route defaults." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to use a compatible configured effort or the selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/text-turn/tool-schemas.expected.json b/snapshots/sdk/text-turn/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/sdk/text-turn/tool-schemas.expected.json +++ b/snapshots/sdk/text-turn/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/tool-error-details/system-prompt.expected.md b/snapshots/sdk/tool-error-details/system-prompt.expected.md index 4dcf8262a6..49a6727bc7 100644 --- a/snapshots/sdk/tool-error-details/system-prompt.expected.md +++ b/snapshots/sdk/tool-error-details/system-prompt.expected.md @@ -173,11 +173,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -206,7 +206,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -214,11 +214,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/sdk/tool-error-details/tool-schemas.expected.json b/snapshots/sdk/tool-error-details/tool-schemas.expected.json index 77b5236350..628380a1d5 100644 --- a/snapshots/sdk/tool-error-details/tool-schemas.expected.json +++ b/snapshots/sdk/tool-error-details/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/advanced-toolchain/system-prompt.expected.md b/snapshots/session/advanced-toolchain/system-prompt.expected.md index 14d8e2559d..2cd6250c0f 100644 --- a/snapshots/session/advanced-toolchain/system-prompt.expected.md +++ b/snapshots/session/advanced-toolchain/system-prompt.expected.md @@ -186,11 +186,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -219,7 +219,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -227,11 +227,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/advanced-toolchain/tool-schemas.expected.json b/snapshots/session/advanced-toolchain/tool-schemas.expected.json index f8cbf6b810..89dd15e503 100644 --- a/snapshots/session/advanced-toolchain/tool-schemas.expected.json +++ b/snapshots/session/advanced-toolchain/tool-schemas.expected.json @@ -303,17 +303,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -397,7 +397,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -415,15 +415,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -1045,17 +1045,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -1139,7 +1139,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -1157,15 +1157,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/agent-instructions/tool-schemas.expected.json b/snapshots/session/agent-instructions/tool-schemas.expected.json index 5018eac594..ffcb7cdbcb 100644 --- a/snapshots/session/agent-instructions/tool-schemas.expected.json +++ b/snapshots/session/agent-instructions/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -924,17 +924,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -1018,7 +1018,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -1036,15 +1036,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/background-confinement-failure/tool-schemas.expected.json b/snapshots/session/background-confinement-failure/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/background-confinement-failure/tool-schemas.expected.json +++ b/snapshots/session/background-confinement-failure/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/both-mode-turn/system-prompt.expected.md b/snapshots/session/both-mode-turn/system-prompt.expected.md index 69bc221437..da16df0b27 100644 --- a/snapshots/session/both-mode-turn/system-prompt.expected.md +++ b/snapshots/session/both-mode-turn/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json index 763c4962d9..f96956dd76 100644 --- a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json @@ -302,17 +302,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -463,7 +463,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -481,15 +481,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json index f8db659ff3..2fc2120472 100644 --- a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json @@ -302,17 +302,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -465,7 +465,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -483,15 +483,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json index f3d955b641..d8f286c03c 100644 --- a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json +++ b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -569,7 +569,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -587,15 +587,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/claude-code-mods/tool-schemas.expected.json b/snapshots/session/claude-code-mods/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/claude-code-mods/tool-schemas.expected.json +++ b/snapshots/session/claude-code-mods/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/compaction-recovery/tool-schemas.expected.json b/snapshots/session/compaction-recovery/tool-schemas.expected.json index 5018eac594..ffcb7cdbcb 100644 --- a/snapshots/session/compaction-recovery/tool-schemas.expected.json +++ b/snapshots/session/compaction-recovery/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -924,17 +924,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -1018,7 +1018,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -1036,15 +1036,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json index 1107b15b8f..4a4cca52f0 100644 --- a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json @@ -302,17 +302,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -442,7 +442,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -460,15 +460,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json index 0bf5c52da2..3957dbc2f8 100644 --- a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json @@ -308,17 +308,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -402,7 +402,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -420,15 +420,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json index e8770b87d3..2e5f4fc018 100644 --- a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json +++ b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json @@ -303,17 +303,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -397,7 +397,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -415,15 +415,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json index de74610301..798fc8024b 100644 --- a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json +++ b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -924,17 +924,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -1026,7 +1026,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -1044,15 +1044,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -1593,17 +1593,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -1687,7 +1687,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -1705,15 +1705,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json +++ b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json index 3fb11acd59..03764bcd7e 100644 --- a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json +++ b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json @@ -182,17 +182,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -260,7 +260,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -278,15 +278,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/lsp-definition/tool-schemas.expected.json b/snapshots/session/lsp-definition/tool-schemas.expected.json index c9da49006c..2bc62b0f70 100644 --- a/snapshots/session/lsp-definition/tool-schemas.expected.json +++ b/snapshots/session/lsp-definition/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -393,7 +393,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -411,15 +411,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md index da3f0b8ab7..b81f602015 100644 --- a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md +++ b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md @@ -162,11 +162,11 @@ interface ToolArgsMap { /** Continuation cursor returned by this server. */ cursor?: string; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -202,7 +202,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -210,11 +210,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/mcp-resources/tool-schemas.expected.json b/snapshots/session/mcp-resources/tool-schemas.expected.json index b7e47f91c6..fa72eb6521 100644 --- a/snapshots/session/mcp-resources/tool-schemas.expected.json +++ b/snapshots/session/mcp-resources/tool-schemas.expected.json @@ -302,17 +302,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -417,7 +417,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -435,15 +435,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json index 8cfd4f8c9f..fe1f0ce7a4 100644 --- a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json +++ b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -375,7 +375,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -393,15 +393,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/office-progressive-native/tool-schemas.expected.json b/snapshots/session/office-progressive-native/tool-schemas.expected.json index f43f78ab5f..36be4284f4 100644 --- a/snapshots/session/office-progressive-native/tool-schemas.expected.json +++ b/snapshots/session/office-progressive-native/tool-schemas.expected.json @@ -217,17 +217,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -311,7 +311,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -329,15 +329,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/office-progressive-ptc/system-prompt.expected.md b/snapshots/session/office-progressive-ptc/system-prompt.expected.md index c5fc982e04..e9119ddc55 100644 --- a/snapshots/session/office-progressive-ptc/system-prompt.expected.md +++ b/snapshots/session/office-progressive-ptc/system-prompt.expected.md @@ -119,11 +119,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -152,7 +152,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -160,11 +160,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/office-skills/tool-schemas.expected.json b/snapshots/session/office-skills/tool-schemas.expected.json index f43f78ab5f..36be4284f4 100644 --- a/snapshots/session/office-skills/tool-schemas.expected.json +++ b/snapshots/session/office-skills/tool-schemas.expected.json @@ -217,17 +217,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -311,7 +311,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -329,15 +329,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json index eaefa6b809..4195c5187a 100644 --- a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json @@ -137,17 +137,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -231,7 +231,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -249,15 +249,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json index c06f67618d..ca83d406e0 100644 --- a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -417,7 +417,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -435,15 +435,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ @@ -1025,17 +1025,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -1210,7 +1210,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -1228,15 +1228,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/plugin-manager/tool-schemas.expected.json b/snapshots/session/plugin-manager/tool-schemas.expected.json index 6357a8beab..62bdec5dc1 100644 --- a/snapshots/session/plugin-manager/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -417,7 +417,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -435,15 +435,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/product-subagent-both/tool-schemas.expected.json b/snapshots/session/product-subagent-both/tool-schemas.expected.json index ad221ee080..08ca2b1a53 100644 --- a/snapshots/session/product-subagent-both/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-both/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/product-subagent-codex/tool-schemas.expected.json b/snapshots/session/product-subagent-codex/tool-schemas.expected.json index 7df665c825..49f6480eb1 100644 --- a/snapshots/session/product-subagent-codex/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-codex/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json index 50fbaad047..34004323e2 100644 --- a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/provider-cwd/tool-schemas.expected.json b/snapshots/session/provider-cwd/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/provider-cwd/tool-schemas.expected.json +++ b/snapshots/session/provider-cwd/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ptc-node-read-only/system-prompt.expected.md b/snapshots/session/ptc-node-read-only/system-prompt.expected.md index 43224a84f3..1e964bdb4f 100644 --- a/snapshots/session/ptc-node-read-only/system-prompt.expected.md +++ b/snapshots/session/ptc-node-read-only/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/ptc-node-workspace/system-prompt.expected.md b/snapshots/session/ptc-node-workspace/system-prompt.expected.md index 43224a84f3..1e964bdb4f 100644 --- a/snapshots/session/ptc-node-workspace/system-prompt.expected.md +++ b/snapshots/session/ptc-node-workspace/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/ptc-persistent/system-prompt.expected.md b/snapshots/session/ptc-persistent/system-prompt.expected.md index 43224a84f3..1e964bdb4f 100644 --- a/snapshots/session/ptc-persistent/system-prompt.expected.md +++ b/snapshots/session/ptc-persistent/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/ptc-policy-change/system-prompt.expected.md b/snapshots/session/ptc-policy-change/system-prompt.expected.md index 07099f54b6..65c682d835 100644 --- a/snapshots/session/ptc-policy-change/system-prompt.expected.md +++ b/snapshots/session/ptc-policy-change/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/ptc-python-turn/system-prompt.expected.md b/snapshots/session/ptc-python-turn/system-prompt.expected.md index 45143cc82e..a6dfe8d41b 100644 --- a/snapshots/session/ptc-python-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-python-turn/system-prompt.expected.md @@ -288,9 +288,9 @@ class ListAgentsOutput2(TypedDict): depth: NotRequired[float] class ListSubagentModelsArgs(TypedDict): - # Registered LLM provider id. Omit to list providers. + # Registered LLM provider id. provider: NotRequired[str] - # Exact model id to inspect. Requires provider; omit to list that provider's allowed models. + # Exact model id; requires provider. model: NotRequired[str] # Additional keys beyond those declared are allowed. @@ -377,11 +377,11 @@ class SubagentArgs(TypedDict): description: str # The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. prompt: str - # LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. + # Child LLM provider id. provider: NotRequired[str] - # Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. + # Model id within that provider. model: NotRequired[str] - # Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. + # Model-specific reasoning effort id. reasoning_effort: NotRequired[str] # Additional keys beyond those declared are allowed. @@ -549,7 +549,7 @@ class Tools(Protocol): async def list_agents(self, args: ListAgentsArgs) -> list[ListAgentsOutput1 | ListAgentsOutput2]: """List subagents you started, with their ids, labels, and status. running means it is working; inactive means it is not currently working. You will be notified when a subagent finishes; there is no need to keep checking its status. Use send_message to continue the conversation.""" async def list_subagent_models(self, args: ListSubagentModelsArgs) -> str: - """Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.""" + """Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.""" async def read(self, args: ReadArgs) -> ReadOutput: """Read a UTF-8 text file and return line-numbered content.""" async def read_image(self, args: ReadImageArgs) -> ReadImageOutput: @@ -559,7 +559,7 @@ class Tools(Protocol): async def skill(self, args: SkillArgs) -> SkillOutput: """Load the full instructions for a skill. Call it before acting on a task that names or clearly matches a skill in the session skill catalog.""" async def subagent(self, args: SubagentArgs) -> SubagentOutput: - """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.""" + """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.""" async def subagent_fork(self, args: SubagentForkArgs) -> SubagentForkOutput: """Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.""" async def todo_write(self, args: TodoWriteArgs) -> TodoWriteOutput: diff --git a/snapshots/session/ptc-read-image/system-prompt.expected.md b/snapshots/session/ptc-read-image/system-prompt.expected.md index 0af031b0c8..4d6f11b10c 100644 --- a/snapshots/session/ptc-read-image/system-prompt.expected.md +++ b/snapshots/session/ptc-read-image/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/ptc-turn/system-prompt.expected.md b/snapshots/session/ptc-turn/system-prompt.expected.md index 43224a84f3..1e964bdb4f 100644 --- a/snapshots/session/ptc-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-turn/system-prompt.expected.md @@ -140,11 +140,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Read a UTF-8 text file and return line-numbered content. */ @@ -173,7 +173,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,11 +181,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json index 816b7fcb6a..9127b4826a 100644 --- a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json +++ b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json index f9cd850b97..c35e8cec15 100644 --- a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json @@ -189,17 +189,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -312,7 +312,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -330,15 +330,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.1.expected.json b/snapshots/session/ralph-loop/tool-schemas.1.expected.json index 0bb4c382a8..d72f9a2ae8 100644 --- a/snapshots/session/ralph-loop/tool-schemas.1.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -419,7 +419,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -437,15 +437,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.2.expected.json b/snapshots/session/ralph-loop/tool-schemas.2.expected.json index 0bb4c382a8..d72f9a2ae8 100644 --- a/snapshots/session/ralph-loop/tool-schemas.2.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.2.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -419,7 +419,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -437,15 +437,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.expected.json b/snapshots/session/ralph-loop/tool-schemas.expected.json index f67a9cc249..9fba724dcc 100644 --- a/snapshots/session/ralph-loop/tool-schemas.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -376,7 +376,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -394,15 +394,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/session-query-spill/tool-schemas.expected.json b/snapshots/session/session-query-spill/tool-schemas.expected.json index 18b7f4a24d..1c778ec2c9 100644 --- a/snapshots/session/session-query-spill/tool-schemas.expected.json +++ b/snapshots/session/session-query-spill/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -560,7 +560,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -578,15 +578,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json index 233a8b87a8..48a1ac30ce 100644 --- a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json index ac4d94f100..b4fa90e046 100644 --- a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json +++ b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json @@ -325,17 +325,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -419,7 +419,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -437,15 +437,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json index 045312a524..dd16436c73 100644 --- a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json @@ -176,17 +176,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -249,7 +249,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -267,15 +267,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json index 9c47c0c649..16a256b818 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -371,7 +371,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -389,15 +389,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json index 2eaec5d921..498673be8a 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json index f9cbf2b0d8..4442bcf828 100644 --- a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json @@ -176,17 +176,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -270,7 +270,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -288,15 +288,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json +++ b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/team-targets/tool-schemas.expected.json b/snapshots/session/team-targets/tool-schemas.expected.json index 3409e4753a..f319d5da9f 100644 --- a/snapshots/session/team-targets/tool-schemas.expected.json +++ b/snapshots/session/team-targets/tool-schemas.expected.json @@ -253,17 +253,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } diff --git a/snapshots/session/text-turn/tool-schemas.expected.json b/snapshots/session/text-turn/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/text-turn/tool-schemas.expected.json +++ b/snapshots/session/text-turn/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/web-fetch/tool-schemas.expected.json b/snapshots/session/web-fetch/tool-schemas.expected.json index 40cc9548ef..114d460759 100644 --- a/snapshots/session/web-fetch/tool-schemas.expected.json +++ b/snapshots/session/web-fetch/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/workflow-confinement/tool-schemas.expected.json b/snapshots/session/workflow-confinement/tool-schemas.expected.json index 8dc90581f4..185fd4ef83 100644 --- a/snapshots/session/workflow-confinement/tool-schemas.expected.json +++ b/snapshots/session/workflow-confinement/tool-schemas.expected.json @@ -262,17 +262,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -374,15 +374,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/workspace-dependencies/tool-schemas.expected.json b/snapshots/session/workspace-dependencies/tool-schemas.expected.json index 9bd9bd41b5..f009027b9e 100644 --- a/snapshots/session/workspace-dependencies/tool-schemas.expected.json +++ b/snapshots/session/workspace-dependencies/tool-schemas.expected.json @@ -217,17 +217,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -319,7 +319,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -337,15 +337,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/worktree/tool-schemas.expected.json b/snapshots/session/worktree/tool-schemas.expected.json index dd66d0a51f..d95e52e13e 100644 --- a/snapshots/session/worktree/tool-schemas.expected.json +++ b/snapshots/session/worktree/tool-schemas.expected.json @@ -279,17 +279,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -373,7 +373,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -391,15 +391,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/web/fresh-round-trip/tool-schemas.expected.json b/snapshots/web/fresh-round-trip/tool-schemas.expected.json index 26471f7526..db2840f35d 100644 --- a/snapshots/web/fresh-round-trip/tool-schemas.expected.json +++ b/snapshots/web/fresh-round-trip/tool-schemas.expected.json @@ -325,17 +325,17 @@ }, { "name": "list_subagent_models", - "description": "Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", "parameters": { "type": "object", "properties": { "provider": { "type": "string", - "description": "Registered LLM provider id. Omit to list providers." + "description": "Registered LLM provider id." }, "model": { "type": "string", - "description": "Exact model id to inspect. Requires provider; omit to list that provider's allowed models." + "description": "Exact model id; requires provider." } } } @@ -719,7 +719,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { @@ -737,15 +737,15 @@ }, "provider": { "type": "string", - "description": "LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route." + "description": "Child LLM provider id." }, "model": { "type": "string", - "description": "Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route." + "description": "Model id within that provider." }, "reasoning_effort": { "type": "string", - "description": "Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default." + "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/web/official-subagent-bundle/system-prompt.expected.md b/snapshots/web/official-subagent-bundle/system-prompt.expected.md index dd702b72a4..6d483f1bb7 100644 --- a/snapshots/web/official-subagent-bundle/system-prompt.expected.md +++ b/snapshots/web/official-subagent-bundle/system-prompt.expected.md @@ -158,11 +158,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied. */ @@ -288,7 +288,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -296,11 +296,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. */ diff --git a/snapshots/web/optional-string-editor/system-prompt.expected.md b/snapshots/web/optional-string-editor/system-prompt.expected.md index 597ebe90cd..5cb0f37436 100644 --- a/snapshots/web/optional-string-editor/system-prompt.expected.md +++ b/snapshots/web/optional-string-editor/system-prompt.expected.md @@ -158,11 +158,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied. */ @@ -305,7 +305,7 @@ interface ToolArgsMap { /** Optional parameter of `view` command when `path` points to a file. If omitted or null, the full file is shown. If provided, the file will be shown in the indicated line number range, e.g. [11, 12] will show lines 11 and 12. Indexing at 1 to start. Setting `[start_line, -1]` shows all lines from `start_line` to the end of the file. */ view_range?: number[] | null; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -313,11 +313,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ diff --git a/snapshots/web/ptc-round/system-prompt.expected.md b/snapshots/web/ptc-round/system-prompt.expected.md index 731d7cc7b4..b5d6fc4177 100644 --- a/snapshots/web/ptc-round/system-prompt.expected.md +++ b/snapshots/web/ptc-round/system-prompt.expected.md @@ -158,11 +158,11 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Discover LLM routes for subagents without changing the current Agent. Call with no arguments to list registered providers, with `provider` to list its allowed models, or with `provider` and `model` to inspect that exact model and its reasoning efforts. Catalog membership is advisory: an adapter may accept an unlisted model id. Use the returned ids with a delegation tool's `provider`, `model`, and `reasoning_effort` fields. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ list_subagent_models: { - /** Registered LLM provider id. Omit to list providers. */ + /** Registered LLM provider id. */ provider?: string; - /** Exact model id to inspect. Requires provider; omit to list that provider's allowed models. */ + /** Exact model id; requires provider. */ model?: string; } & Record; /** Declare existing files as final deliverables for the user. Use it when the user needs a separate file, especially Office documents, spreadsheets, and slide decks; prefer your final response when that suffices. The user opens the current files; their contents are not copied. */ @@ -288,7 +288,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Child LLM selection is optional. Omit `provider`, `model`, and `reasoning_effort` to use configured child defaults and inherit compatible missing values from the parent Agent. Supply `provider` and `model` together after using `list_subagent_models` to inspect advertised routes and efforts. Changing the effective route without naming an effort uses the selected model's default effort. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -296,11 +296,11 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** LLM provider route for the child. Supply together with model; omit both to use configured child defaults or inherit the parent route. */ + /** Child LLM provider id. */ provider?: string; - /** Model id interpreted by provider. Supply together with provider; omit both to use configured child defaults or inherit the parent route. */ + /** Model id within that provider. */ model?: string; - /** Adapter-owned reasoning effort for the effective child route. Omit to inherit a compatible configured/parent effort or use a newly selected model's default. */ + /** Model-specific reasoning effort id. */ reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ From 764a51d5071855871ae917a4cc2f433c0d3d68da Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Mon, 12 Oct 2026 18:37:26 +0800 Subject: [PATCH 016/162] refactor(subagent): expose model selection only on settings-owning tools Route fields and list_subagent_models are now registered only by the tool-subagent instance configured with modelSelectionSettings, which keeps one fixed definition per composition for Web presets while headless and SDK profiles return to their previous schemas. Discovery registers with that instance instead of a separate list-models plugin entry, so the delegation description never references an absent tool. modelSelectionSettings on a provider that inherits the parent conversation now fails the mount. --- ...authorized-subagent-model-routes.i18n.yaml | 8 +- ...4-user-authorized-subagent-model-routes.md | 4 +- ...ser-authorized-subagent-model-routes.zh.md | 4 +- apps/cli/composition.md | 3 - .../parent.expected.jsonl | 4 +- .../tests/profiles/sdk/keyless-smoke.e2e.ts | 2 +- .../plugin-instances.expected.md | 4 - docs/config-catalog.md | 9 +- docs/config-catalog.zh.md | 9 +- docs/tool-catalog.i18n.yaml | 8 +- docs/tool-catalog.md | 18 +--- docs/tool-catalog.zh.md | 18 +--- .../guide.i18n.yaml | 16 ++++ .../fork-model-selection-settings/guide.md | 28 ++++++ .../fork-model-selection-settings/guide.zh.md | 28 ++++++ packages/bundle/base/cordis.patch.yml | 3 - packages/bundle/web-app/cordis.patch.yml | 3 - .../bundle/web-app/presets/cordis.patch.yml | 2 - packages/bundle/web-app/presets/ptc.patch.yml | 2 - .../bundle/web-app/presets/standard.patch.yml | 2 - .../subagent/tool-subagent/README.i18n.yaml | 32 +++---- packages/subagent/tool-subagent/README.md | 19 ++-- packages/subagent/tool-subagent/README.zh.md | 19 ++-- packages/subagent/tool-subagent/package.json | 5 - packages/subagent/tool-subagent/src/index.ts | 27 ++++-- .../subagent/tool-subagent/src/list-models.ts | 23 +---- .../tests/fixtures/model-discovery.mjs | 38 -------- .../subagent/tool-subagent/tests/harness.ts | 9 +- .../tool-subagent/tests/list-models.spec.ts | 30 ++++-- .../tests/model-selection-settings.spec.ts | 4 - .../tests/model-selection.spec.ts | 48 ++++++++-- .../tool-subagent/tests/tool-subagent.spec.ts | 3 - .../subagent/tool-subagent/tsdown.config.ts | 1 - scripts/check-workspace-constraints.ts | 5 +- scripts/gen-tool-catalog.ts | 6 +- .../advanced/result.json | 6 -- .../advanced/session.1.v4.jsonl | 2 +- .../advanced/session.2.v4.jsonl | 2 +- .../advanced/session.v4.jsonl | 4 +- .../dynamic-tools/tool-history.json | 3 - .../restart/requests.json | 2 - .../restart/session.1.v4.jsonl | 2 +- .../restart/session.2.v4.jsonl | 2 +- .../scheduler-recovery/session.v4.jsonl | 2 +- .../sdk/bash-tool/tool-schemas.expected.json | 31 +------ .../notifications.expected.jsonl | 6 +- .../sdk/dynamic-tool-updates/session.v4.jsonl | 6 +- snapshots/sdk/persistent-tools/cordis.yml | 4 - .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.expected.json | 59 +----------- .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.1.expected.json | 31 +------ .../sdk/text-turn/tool-schemas.expected.json | 31 +------ .../notifications.expected.jsonl | 4 +- .../sdk/tool-error-details/session.v4.jsonl | 4 +- .../system-prompt.expected.md | 16 +--- .../tool-schemas.expected.json | 31 +------ .../advanced-toolchain/session.v4.jsonl | 8 +- .../system-prompt.expected.md | 16 +--- .../tool-schemas.expected.json | 62 +------------ .../tool-schemas.expected.json | 62 +------------ .../tool-schemas.expected.json | 31 +------ .../both-mode-turn/system-prompt.expected.md | 16 +--- .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../cordis.snapshot.yml | 4 +- .../compaction-output-reserve/cordis.yml | 2 +- .../session.v4.jsonl | 2 +- .../tool-schemas.expected.json | 62 +------------ .../cordis.snapshot.yml | 4 +- .../compaction-summary-headroom/cordis.yml | 2 +- .../session.v4.jsonl | 2 +- .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../session.v4.jsonl | 6 +- .../dynamic-tool-updates/session.v4.jsonl | 6 +- .../tool-schemas.expected.json | 93 +------------------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../lsp-definition/tool-schemas.expected.json | 31 +------ .../system-prompt.expected.md | 16 +--- .../mcp-resources/tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../system-prompt.expected.md | 16 +--- .../office-skills/tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../session.v4.jsonl | 6 +- .../plugin-manager-mcp/session.v4.jsonl | 4 +- .../tool-schemas.expected.json | 62 +------------ .../plugin-manager/tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../provider-cwd/tool-schemas.expected.json | 31 +------ .../system-prompt.expected.md | 16 +--- .../system-prompt.expected.md | 16 +--- .../ptc-persistent/system-prompt.expected.md | 16 +--- .../system-prompt.expected.md | 16 +--- .../ptc-python-turn/system-prompt.expected.md | 17 +--- .../ptc-read-image/system-prompt.expected.md | 16 +--- .../ptc-turn/system-prompt.expected.md | 16 +--- .../tool-schemas.expected.json | 31 +------ .../pwsh-tool-turn/tool-schemas.expected.json | 31 +------ .../ralph-loop/tool-schemas.1.expected.json | 31 +------ .../ralph-loop/tool-schemas.2.expected.json | 31 +------ .../ralph-loop/tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../runtime.cordis.yml | 3 - .../session.v4.jsonl | 23 ----- .../snapshot.yml | 7 -- .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.1.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../team-targets/tool-schemas.expected.json | 17 ---- .../text-turn/tool-schemas.expected.json | 31 +------ .../web-fetch/tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../tool-schemas.expected.json | 31 +------ .../worktree/tool-schemas.expected.json | 31 +------ tsconfig.base.json | 1 - 133 files changed, 353 insertions(+), 2380 deletions(-) create mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml create mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md create mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md delete mode 100644 packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs delete mode 100644 snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml delete mode 100644 snapshots/session/subagent-model-discovery-errors/session.v4.jsonl delete mode 100644 snapshots/session/subagent-model-discovery-errors/snapshot.yml diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index 2592a568ad..250e5e27cf 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -9,14 +9,14 @@ en: 3238dfb7db7dfdd8 zh: ec3bdd9a9da0a321 /agent-note-user-authorized-subagent-model-routes/decision: - en: eba9e531eb6c450b - zh: 7c0673644e83608a + en: e67644944167a865 + zh: dc2b28f630a0f6fa /agent-note-user-authorized-subagent-model-routes/alternatives-considered: en: e41c17226190ea8b zh: d3b1395fafe75f01 /agent-note-user-authorized-subagent-model-routes/consequences: - en: 1749cfbfda849692 - zh: 9925e638cd72fba6 + en: 36b76692f1605e6d + zh: a63a249c980bc654 /agent-note-user-authorized-subagent-model-routes/related-decisions: en: 0eead87ec1172d81 zh: 4e776a583f5a00a5 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index a91f0e9e47..357ef78e36 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -16,7 +16,7 @@ A newly composed top-level Session snapshots the route list in `subagent/model-s The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and intersects their advertised model catalogs with the recorded route list. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts, including for ids absent from the catalog. Without a recorded policy, discovery resolves only the parent Agent’s current effective route, even if the catalog omits it. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. -Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. The primary spawn tool captures the setting once, while all capable delegation tools and shared discovery read the same Session policy. Fork tools reject model-selection arguments to preserve inherited-prefix reuse. +Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. Only a delegation tool configured with `modelSelectionSettings` exposes the route fields and registers `list_subagent_models`; the shipped Web spawn tool is that owner, captures the setting once, and reads the Session policy at each call. Fork tools reject model-selection arguments, and the setting fails their mount, to preserve inherited-prefix reuse. ## Alternatives considered @@ -34,7 +34,7 @@ Model selection has no unrestricted static mode. The default-off Host setting au - New adapter registrations and newly advertised models do not expand user authorization. - Adapter removal or catalog lookup failure can make provider discovery unavailable without deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. -- The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. +- The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. Compositions with `modelSelectionSettings` carry the three route fields and the discovery schema in every Session; headless and SDK profiles carry neither. - The policy event is log-only and is appended while an Agent is composed, before either SDK begins its run subscription. Shipped SDK profiles do not enable this Web-owned preference, so the event changes neither SDK's expected notifications or persisted-session output; package restore tests own its durable projection instead of fabricating an SDK composition solely to emit it. - Unit coverage pins settings validation, malformed durable values, Session sampling and inheritance, shared definitions, parent-route defaults, discovery restrictions, executor denial, and the settings UI’s staged writes and revision checks. The assembled Web scenario pins the real settings document and Plugins card flow. diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index 49a221d220..4fe117478a 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -16,7 +16,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` 固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并取其公布模型目录与已记录路由列表的交集。精确查询先要求授权，再解析模型元数据及全部支持的推理强度，也允许目录未列出的 ID。没有已记录策略时，发现只解析父 Agent 当前生效的路由，即使目录未列出它。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 -模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。主 spawn 工具一次性记录设置，所有支持选择的委派工具和共享发现均读取同一 Session 策略。fork 工具拒绝模型选择参数，以保留对继承前缀的复用。 +模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。只有配置了 `modelSelectionSettings` 的委派工具会公开路由字段并注册 `list_subagent_models`；随附 Web 的 spawn 工具就是该所有者，它一次性记录设置，并在每次调用时读取 Session 策略。fork 工具拒绝模型选择参数，且该设置会使其挂载失败，以保留对继承前缀的复用。 ## Alternatives considered @@ -34,7 +34,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` - 新适配器注册和新公布模型不会扩大用户授权。 - 适配器移除或目录查询失败可能使提供方发现不可用，但不会删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 -- 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。 +- 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。配置了 `modelSelectionSettings` 的组合在每个 Session 中都携带三个路由字段和发现 schema；headless 与 SDK profile 两者都不携带。 - 策略事件仅存在于日志，并在 Agent 组合期间、两套 SDK 开始订阅运行前追加。随附 SDK profile 不启用这项 Web 自有偏好，因此该事件不会改变任一 SDK 的预期通知或持久 Session 输出；其持久投影由包级恢复测试负责，不会为了发出该事件而虚构 SDK 组合。 - 单元覆盖固定设置校验、异常持久值、Session 取样与继承、共享定义、父级路由默认值、发现限制、执行器拒绝，以及设置 UI 的暂存写入与 revision 检查。组装 Web 场景固定真实设置文档与 Plugins 设置卡流程。 diff --git a/apps/cli/composition.md b/apps/cli/composition.md index f2bae814f4..9888602296 100644 --- a/apps/cli/composition.md +++ b/apps/cli/composition.md @@ -144,8 +144,6 @@ flowchart LR cfg --> plugin_dsh_base_subagent_fork_in_process plugin_dsh_base_tool_subagent_control["tool-subagent-control
@deepseek-ai/dsh-tool-subagent-control"] cfg --> plugin_dsh_base_tool_subagent_control - plugin_dsh_base_tool_subagent_list_models["tool-subagent-list-models
@deepseek-ai/dsh-tool-subagent/list-models"] - cfg --> plugin_dsh_base_tool_subagent_list_models plugin_dsh_base_tool_subagent_list_agents["tool-subagent-list-agents
@deepseek-ai/dsh-tool-subagent-control/list-agents"] cfg --> plugin_dsh_base_tool_subagent_list_agents plugin_dsh_base_tool_subagent["tool-subagent
@deepseek-ai/dsh-tool-subagent"] @@ -270,7 +268,6 @@ flowchart LR | `subagent-spawn-in-process` | `@deepseek-ai/dsh-subagent-spawn-in-process` | | `subagent-fork-in-process` | `@deepseek-ai/dsh-subagent-fork-in-process` | | `tool-subagent-control` | `@deepseek-ai/dsh-tool-subagent-control` | -| `tool-subagent-list-models` | `@deepseek-ai/dsh-tool-subagent/list-models` | | `tool-subagent-list-agents` | `@deepseek-ai/dsh-tool-subagent-control/list-agents` | | `tool-subagent` | `@deepseek-ai/dsh-tool-subagent` | | `tool-subagent-fork` | `@deepseek-ai/dsh-tool-subagent` | diff --git a/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl b/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl index 01674e2a6e..1cee326a29 100644 --- a/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl +++ b/apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl @@ -17,8 +17,8 @@ {"type":"system/message","data":{"turn":2,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"sourceEventSeqs":[2],"surfaceOp":{"op":"replace","startSeq":2,"endSeq":2}} {"type":"user/message","data":{"content":[{"type":"text","text":"Delegate the write probe to a subagent."}],"source":{"kind":"user"},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: read-only. Any available operation enforced by the DSH file sandbox cannot modify files in the standing mode. Do not refuse a required modification from this policy alone: try an available tool normally and follow any denial and escalation guidance it returns.\n\nApproval policy: ask. Operations that require approval may ask through the configured answerers; without an available answerer, the request fails closed."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: read-only. Any available operation enforced by the DSH file sandbox cannot modify files in the standing mode. Do not refuse a required modification from this policy alone: try an available tool normally and follow any denial and escalation guidance it returns."},{"name":"approval:policy","text":"Approval policy: ask. Operations that require approval may ask through the configured answerers; without an available answerer, the request fails closed."}]},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","reasoningEffort":"low"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"resume","startsSeries":true}} -{"type":"developer/message","data":{"turn":2,"step":1,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"list_subagent_models"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":18},"surfaceOp":"append"} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","reasoningEffort":"low"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"resume","startsSeries":true}} +{"type":"developer/message","data":{"turn":2,"step":1,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":18},"surfaceOp":"append"} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Tighten this session to read-only.","messageSeqs":[3],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":2,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"delegate-write","name":"subagent","arguments":"{\"description\": \"Delegated write probe\", \"prompt\": \"Use the write tool exactly once with file_path set to exactly the relative path inherited.txt and content escaped. If the write is denied, reply with the single word CHILD_DENIED and the denial marker line; do not retry and do not request escalation. If it succeeds, reply CHILD_WROTE.\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"delegate-write","name":"subagent","args":["{\"description\": \"Delegated write probe\", \"prompt\": \"Use the write tool exactly once with file_path set to exactly the relative path inherited.txt and content escaped. If the write is denied, reply with the single word CHILD_DENIED and the denial marker line; do not retry and do not request escalation. If it succeeds, reply CHILD_WROTE.\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"delegate-write","name":"subagent","arguments":"{\"description\": \"Delegated write probe\", \"prompt\": \"Use the write tool exactly once with file_path set to exactly the relative path inherited.txt and content escaped. If the write is denied, reply with the single word CHILD_DENIED and the denial marker line; do not retry and do not request escalation. If it succeeds, reply CHILD_WROTE.\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} diff --git a/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts b/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts index 5ee1288135..d0f44a6356 100644 --- a/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts +++ b/apps/cli/tests/profiles/sdk/keyless-smoke.e2e.ts @@ -178,7 +178,7 @@ describe('Python SDK dsh profile keyless smoke', () => { expect(modelRequests[0]?.max_tokens).toBe(1234) expect(toolNames).toEqual(expect.arrayContaining(['read', 'write', 'edit', 'web_fetch', 'web_search'])) expect(toolNames.includes('str_replace_editor')).toBe(editorEnabled) - expect(toolNames).toContain('list_subagent_models') + expect(toolNames).not.toContain('list_subagent_models') child.stdin.write(`${JSON.stringify({ jsonrpc: '2.0', id: 3, method: 'shutdown' })}\n`) const shutdown = await waitForLine(lines, value => value.id === 3, () => stderr) diff --git a/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md b/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md index 9c455c27da..239b4855e7 100644 --- a/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md +++ b/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md @@ -4,10 +4,6 @@ - strong: 子智能体控制工具 - text: 让 AI 向子智能体发送消息或中断其运行。 - code: tool-subagent-control - - listitem: - - button "tool-subagent/list-models, tool-subagent-list-models, 已启用": - - strong: tool-subagent/list-models - - code: tool-subagent-list-models - listitem: - button "子智能体列表工具, tool-subagent-list-agents, 已启用": - strong: 子智能体列表工具 diff --git a/docs/config-catalog.md b/docs/config-catalog.md index a40560130b..48f63cf574 100644 --- a/docs/config-catalog.md +++ b/docs/config-catalog.md @@ -3977,7 +3977,7 @@ export interface Config { - `inject`: `tools` · `subagents` · `systemPrompt` · `sessionProjections` - `refs`: [`AgentOptions`](subsystems/core.md) -- `source`: [`packages/subagent/tool-subagent/src/index.ts:47`](../packages/subagent/tool-subagent/src/index.ts) +- `source`: [`packages/subagent/tool-subagent/src/index.ts:48`](../packages/subagent/tool-subagent/src/index.ts) ```ts config-catalog /** Config: which registered provider this tool delegates to, plus child defaults. */ @@ -3990,8 +3990,11 @@ export interface Config { */ toolName?: string /** - * Sample the Host `subagent-model-selection` setting for each new top-level - * Session and inherit that decision in its child Sessions. + * Expose optional child `provider`, `model`, and `reasoning_effort` fields + * plus `list_subagent_models`, and sample the Host `subagent-model-selection` + * setting for each new top-level Session; child Sessions inherit that + * decision. Requires a provider with `agentOptions` that does not inherit + * the parent conversation. At most one instance per tool scope may enable it. */ modelSelectionSettings?: boolean /** diff --git a/docs/config-catalog.zh.md b/docs/config-catalog.zh.md index 10c01cbe78..b462472c26 100644 --- a/docs/config-catalog.zh.md +++ b/docs/config-catalog.zh.md @@ -3979,7 +3979,7 @@ export interface Config { - `inject`: `tools` · `subagents` · `systemPrompt` · `sessionProjections` - `refs`: [`AgentOptions`](subsystems/core.zh.md) -- `source`: [`packages/subagent/tool-subagent/src/index.ts:47`](../packages/subagent/tool-subagent/src/index.ts) +- `source`: [`packages/subagent/tool-subagent/src/index.ts:48`](../packages/subagent/tool-subagent/src/index.ts) ```ts config-catalog /** Config: which registered provider this tool delegates to, plus child defaults. */ @@ -3992,8 +3992,11 @@ export interface Config { */ toolName?: string /** - * Sample the Host `subagent-model-selection` setting for each new top-level - * Session and inherit that decision in its child Sessions. + * Expose optional child `provider`, `model`, and `reasoning_effort` fields + * plus `list_subagent_models`, and sample the Host `subagent-model-selection` + * setting for each new top-level Session; child Sessions inherit that + * decision. Requires a provider with `agentOptions` that does not inherit + * the parent conversation. At most one instance per tool scope may enable it. */ modelSelectionSettings?: boolean /** diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 4b23c19b5c..529e9be439 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: 559ba7c9e11d4919 - zh: 44c540c14cf43f95 + en: 551b31e8c713bbcd + zh: 861604a6d4d6db6b /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -228,8 +228,8 @@ en: 6872bee11fccb6c8 zh: 230f2d443dc5d9a6 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: aff4a6668b41b7d5 - zh: 98d6ddf8aa28c7f9 + en: d73236d49ab7c9f5 + zh: 1009b5c5d6d678a5 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 769ad76242..090afd911d 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions without inheriting parent context always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` rejects model-selection arguments; without an enabled policy, spawn selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2039,7 +2039,7 @@ Source: [`packages/subagent/tool-subagent/src/list-models.ts`](../packages/subag ### `subagent` -Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. +Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. ```json { @@ -2056,18 +2056,6 @@ Delegate a self-contained task to a subagent (a separate agent that works in its "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -2079,7 +2067,7 @@ Delegate a self-contained task to a subagent (a separate agent that works in its Source: [`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions without inheriting parent context always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` rejects model-selection arguments; without an enabled policy, spawn selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. +The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index a91c05a5cb..b60ade9669 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 且不继承父级上下文的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 拒绝模型选择参数；没有启用策略时，spawn 选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。`subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2045,7 +2045,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， ### `subagent` -将一项自包含任务委派给 subagent（在自身上下文中工作的独立 agent），用它卸载聚焦且独立的工作，例如研究、限定范围的实现或分析，以免消耗当前对话的上下文。subagent 会返回结果，但不会返回中间步骤。此工具启动独立管理的 subagent 并立即返回其 id。运行时会在任务结束时通知你。子代理通过 `send_message` 汇报结果；你可以使用 `send_message` 在运行时引导它，或在结束后继续对话。先通过 `list_subagent_models` 查询选项，再同时设置 `provider` 与 `model`。省略的值使用子级默认值，缺失时使用兼容的父级值。生效路由改变时，省略 `reasoning_effort` 会使用模型的默认强度。 +将一项自包含任务委派给 subagent（在自身上下文中工作的独立 agent），用它卸载聚焦且独立的工作，例如研究、限定范围的实现或分析，以免消耗当前对话的上下文。subagent 会返回结果，但不会返回中间步骤。此工具启动独立管理的 subagent 并立即返回其 id。运行时会在任务结束时通知你。子代理通过 `send_message` 汇报结果；你可以使用 `send_message` 在运行时引导它，或在结束后继续对话。 ```json { @@ -2062,18 +2062,6 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -2085,7 +2073,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， 来源：[`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；支持 agentOptions 且不继承父级上下文的后端始终公开可选路由字段，共享发现插件解析调用 Session 允许的路由。Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；`subagent_fork` 拒绝模型选择参数；没有启用策略时，spawn 选择和发现仅限当前父级路由。每个实例通过 `provider` 选择后端，通过 `modelSelectionSettings` 读取模型选择设置。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 +注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。`subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml new file mode 100644 index 0000000000..c3a98c82ac --- /dev/null +++ b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml @@ -0,0 +1,16 @@ +# Bilingual-pair consistency record for guide.md (docs/i18n/README.md): per heading +# section, a hash of its English and Chinese blocks outside code blocks and generated regions. +# After editing either side, bring the other along and re-record with: +# pnpm run verify-translation-pairing --write docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md +/: + en: 9a987818d6903fcb + zh: 8a4eefd0e7b4ad04 +/remove-modelselectionsettings-from-fork-delegation-tools: + en: f6bf4e381476a12f + zh: 0e745d803fc38fc9 +/remove-modelselectionsettings-from-fork-delegation-tools/change: + en: ccac82df3a00bf69 + zh: 1820a052adff5f6b +/remove-modelselectionsettings-from-fork-delegation-tools/migration: + en: ba409934bc81f492 + zh: fda2f67f731cd1d2 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md new file mode 100644 index 0000000000..542cc9a183 --- /dev/null +++ b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md @@ -0,0 +1,28 @@ +--- +kind: upgrade-guide +description: "`modelSelectionSettings` on a fork delegation tool now fails the mount instead of exposing child route selection." +--- + +# Remove `modelSelectionSettings` from fork delegation tools + +English | [中文](guide.zh.md) + +## Change + +In v0.2.1-alpha.2, a `@deepseek-ai/dsh-tool-subagent` row whose provider inherits the parent conversation, such as the in-process `fork` provider, accepted `modelSelectionSettings: true` and exposed `provider`, `model`, `reasoning_effort`, and `list_subagent_models` in Sessions with an enabled policy. The tool now rejects that configuration when the provider is mounted, so the composition fails to load with `provider "" inherits the parent conversation and does not accept child model selection`. Fork children keep the parent route so the inherited conversation prefix stays eligible for KV Cache reuse. + +The shipped profiles are unaffected: only the Web spawn tool sets `modelSelectionSettings`. A tool row with the setting now exposes the route fields and `list_subagent_models` in every Session; without a recorded Session policy, explicit selection permits only the parent route. + +## Migration + +1. In `cordis.yml`, profile patches, and overlays, remove `modelSelectionSettings` from every `@deepseek-ai/dsh-tool-subagent` row whose `provider` names a fork provider: + + ```yaml + - name: '@deepseek-ai/dsh-tool-subagent' + config: + provider: fork + toolName: subagent_fork + ``` + +2. To let the model choose child routes, set `modelSelectionSettings: true` on one spawn-style row per tool scope instead. +3. Start the composition and confirm that it loads and that the fork tool schema lists only `cwd`, `description`, and `prompt`. diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md new file mode 100644 index 0000000000..e400299473 --- /dev/null +++ b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md @@ -0,0 +1,28 @@ +--- +kind: upgrade-guide +description: "fork 委派工具上的 `modelSelectionSettings` 现在会使挂载失败，而不再公开子级路由选择。" +--- + +# 从 fork 委派工具中移除 `modelSelectionSettings` + +[English](guide.md) | 中文 + +## 变更 + +在 v0.2.1-alpha.2 中，如果 `@deepseek-ai/dsh-tool-subagent` 配置行的提供方会继承父级对话（例如进程内 `fork` 提供方），它会接受 `modelSelectionSettings: true`，并在策略已启用的 Session 中公开 `provider`、`model`、`reasoning_effort` 和 `list_subagent_models`。现在该工具会在提供方挂载时拒绝这一配置，组合加载失败并报告 `provider "" inherits the parent conversation and does not accept child model selection`。fork 子级保持父级路由，使继承的对话前缀继续符合 KV Cache 复用条件。 + +随附 profile 不受影响：只有 Web 的 spawn 工具设置了 `modelSelectionSettings`。设置了该项的工具行现在会在每个 Session 中公开路由字段和 `list_subagent_models`；没有已记录 Session 策略时，显式选择只允许父级路由。 + +## 迁移 + +1. 在 `cordis.yml`、profile patch 和 overlay 中，从所有 `provider` 指向 fork 提供方的 `@deepseek-ai/dsh-tool-subagent` 配置行移除 `modelSelectionSettings`： + + ```yaml + - name: '@deepseek-ai/dsh-tool-subagent' + config: + provider: fork + toolName: subagent_fork + ``` + +2. 如需让模型选择子级路由，请改为在每个工具作用域中的一个 spawn 类配置行上设置 `modelSelectionSettings: true`。 +3. 启动组合，确认它能加载，且 fork 工具的 schema 只列出 `cwd`、`description` 和 `prompt`。 diff --git a/packages/bundle/base/cordis.patch.yml b/packages/bundle/base/cordis.patch.yml index f162113cae..de65cd8015 100644 --- a/packages/bundle/base/cordis.patch.yml +++ b/packages/bundle/base/cordis.patch.yml @@ -369,9 +369,6 @@ - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' - - id: tool-subagent-list-models - name: '@deepseek-ai/dsh-tool-subagent/list-models' - - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' diff --git a/packages/bundle/web-app/cordis.patch.yml b/packages/bundle/web-app/cordis.patch.yml index 18d6983aef..a1ca0b5483 100644 --- a/packages/bundle/web-app/cordis.patch.yml +++ b/packages/bundle/web-app/cordis.patch.yml @@ -552,9 +552,6 @@ - id: tool-subagent-control disabled: true -- id: tool-subagent-list-models - disabled: true - - id: tool-subagent-list-agents disabled: true diff --git a/packages/bundle/web-app/presets/cordis.patch.yml b/packages/bundle/web-app/presets/cordis.patch.yml index e324a4c837..8c8d567ae3 100644 --- a/packages/bundle/web-app/presets/cordis.patch.yml +++ b/packages/bundle/web-app/presets/cordis.patch.yml @@ -87,8 +87,6 @@ config: - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' - - id: tool-subagent-list-models - name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' - id: tool-subagent diff --git a/packages/bundle/web-app/presets/ptc.patch.yml b/packages/bundle/web-app/presets/ptc.patch.yml index ac0cc611ef..8e1ec1fb70 100644 --- a/packages/bundle/web-app/presets/ptc.patch.yml +++ b/packages/bundle/web-app/presets/ptc.patch.yml @@ -88,8 +88,6 @@ config: - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' - - id: tool-subagent-list-models - name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' - id: tool-subagent diff --git a/packages/bundle/web-app/presets/standard.patch.yml b/packages/bundle/web-app/presets/standard.patch.yml index 05b06df618..6a42d0a351 100644 --- a/packages/bundle/web-app/presets/standard.patch.yml +++ b/packages/bundle/web-app/presets/standard.patch.yml @@ -88,8 +88,6 @@ config: - id: tool-subagent-control name: '@deepseek-ai/dsh-tool-subagent-control' - - id: tool-subagent-list-models - name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' - id: tool-subagent diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index fa0bd6c782..a806e98777 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -18,20 +18,20 @@ en: 93f1701f5209ecb5 zh: 6310fd012400aead /deepseek-ai-dsh-tool-subagent/use-this-package/minimal-configuration: - en: d01d4425bc1875dd - zh: f3d36c8a1e5ed0fa + en: 70e715a7262f45b2 + zh: a94e81d7f1513fdf /deepseek-ai-dsh-tool-subagent/use-this-package/managed-delegation: en: 42673064ba88a211 zh: 8fa26285bd649dc9 /deepseek-ai-dsh-tool-subagent/use-this-package/selecting-a-child-llm: - en: f85faf2e965a541a - zh: 3c232e236199b9aa + en: 2bf262d3f462f5f5 + zh: c443c5366f230eee /deepseek-ai-dsh-tool-subagent/understand-the-implementation: en: 6ce5f749256a7e6c zh: a4f2adb04a70d9cf /deepseek-ai-dsh-tool-subagent/understand-the-implementation/design-concept: - en: 5c97aae02694c27a - zh: 5507f580d3e0c43b + en: 794a71280d73dbc2 + zh: acb6990a599f3f66 /deepseek-ai-dsh-tool-subagent/understand-the-implementation/activation-ownership: en: ef692c4d17290920 zh: b1d0d3dab011e2cd @@ -54,11 +54,11 @@ en: e4c7376d4f29dbd5 zh: 6b7a9eaf69c9ea53 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/what-the-model-sees: - en: 56361a0d507bad66 - zh: 979d502dd3ac168f + en: 8266f5b60ac830db + zh: 6dceefd04897b6b0 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/token-effect: - en: 6a1a6351a3a88424 - zh: c574e9ee286518ed + en: 86d1dafcb6c00c57 + zh: fe890e9c6146c45e /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/kv-cache-effect: en: f746a3c2f2f3cd18 zh: 68de857073b229ff @@ -66,11 +66,11 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: be3c5b647252ddc7 - zh: efecd46ed7dd9a95 + en: d0df5c0c9c988408 + zh: cadaa5591dcc5784 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: - en: cfdfe3b1106c3b57 - zh: 3f421f2901d6f7ab + en: cdb02d92cc19062e + zh: c466d4592828021f /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/kv-cache-effect: en: 7dc47a34d5574fbd zh: bd7bd2a4da2f8ae2 @@ -114,8 +114,8 @@ en: f5d5f97a153f7c70 zh: 3d6bc05b9eadfbc2 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work: - en: 897c9481685a6eab - zh: 444ee7ef4d353f60 + en: 8a0d48a82dff3343 + zh: 604ecd4e288ae322 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work/dev-note: en: e67f633dc6fe7773 zh: 3bb0c20c93f16f98 diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index 8591e118e9..6f188766d6 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -36,7 +36,6 @@ Load the subagent service, an in-process or remote backend, and this tool; then ```yaml - name: '@deepseek-ai/dsh-subagent' - name: '@deepseek-ai/dsh-subagent-spawn-in-process' -- name: '@deepseek-ai/dsh-tool-subagent/list-models' - name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn @@ -47,7 +46,7 @@ Load the subagent service, an in-process or remote backend, and this tool; then |---|---|---| | `provider` | required | Provider name on `ctx.subagents` (e.g. `spawn`, `fork`, `acp`) | | `toolName` | `subagent` | Model-facing tool name; distinct for every loaded instance | -| `modelSelectionSettings` | `false` | Sample the Host's exact-route authorization preference for each top-level Session; a standing preset observes matching Sessions, while direct Agent setup passes its Session explicitly; requires provider `agentOptions` support | +| `modelSelectionSettings` | `false` | Expose child route fields and `list_subagent_models`, and sample the Host's exact-route authorization preference for each top-level Session; a standing preset observes matching Sessions, while direct Agent setup passes its Session explicitly; requires provider `agentOptions` support without parent-context inheritance | | `agentOptions` | — | Configured child `provider`, `model`, adapter-owned `reasoningEffort`, and positive `maxTokens` defaults; requires provider `agentOptions` support and overlays any provider-owned route defaults | | `persona` | — | Per-child persona; requires the provider's `persona` capability | | `toolFilter` | — | Per-child global-tool restriction; requires the `toolFilter` capability | @@ -65,7 +64,7 @@ Every backend’s completion notice carries its final answer. Local Spawn and Fo ### Selecting a child LLM -Backends that support `agentOptions` and do not inherit parent context always expose optional `provider`, `model`, and `reasoning_effort` fields. Load `@deepseek-ai/dsh-tool-subagent/list-models` once alongside the delegation tools for shared discovery; shipped compositions include it. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. Set `modelSelectionSettings: true` to sample the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. A restored Session without a policy keeps the parent-route restriction. ACP, Codex, and Claude Code do not support `agentOptions` and reject selection fields. +Set `modelSelectionSettings: true` to expose optional `provider`, `model`, and `reasoning_effort` fields and register `list_subagent_models` with this instance; instances without it expose neither and reject selection fields. The instance samples the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. This mode requires a backend that advertises `agentOptions` and does not inherit the parent conversation; both in-process Spawn and DSH SDK qualify, while Fork, ACP, Codex, and Claude Code fail the mount. A call supplies `provider` and `model` together, or supplies only an effort when configured, parent, or provider-owned defaults provide the route. Static `provider.agentRouteDefaults`, when present, form the provider/model baseline; tool configuration and model fields overlay it before route-aware effort merging and exact-route preflight. Providers without these defaults use compatible values from the parent's latest logged request, then the parent's creation options before its first request, while retaining the configured `maxTokens`. Changing the route without an explicit effort clears the inherited route-owned effort, so the selected model resolves its default. The live LLM adapter validates the effective route before child creation. Catalog membership remains advisory, so a model can use an unlisted id when its adapter accepts it. @@ -81,7 +80,7 @@ This section explains how the tool mirrors provider lifecycle and settles runs; ### Design concept -One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Settings-backed instances capture Session authorization during Agent creation; calls read the durable policy independently of registration. Direct Agent setup supplies its unpublished Session explicitly. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. Discovery has one separate composition entry. No runtime invariant companion is published: policy comes directly from the Session projection and is checked at execution, with no separately retained per-Agent definition state to reconcile. +One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Settings-backed instances also register `list_subagent_models` in the same composition and capture Session authorization during Agent creation; calls read the durable policy independently of registration. Direct Agent setup supplies its unpublished Session explicitly. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. At most one instance in a tool scope may set `modelSelectionSettings` because `list_subagent_models` has a global name. No runtime invariant companion is published: policy comes directly from the Session projection and is checked at execution. ### Activation ownership @@ -129,11 +128,11 @@ Read these pages when the package-level contract is not enough; they move from t #### What the model sees -The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. Backends supporting `agentOptions` that do not inherit parent context always expose `provider`, `model`, and `reasoning_effort`. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. +The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. An instance with `modelSelectionSettings` adds `provider`, `model`, and `reasoning_effort` plus route guidance in every Session. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. #### Token effect -Fixed schema cost per parent request, including three optional route fields on capable backends. Each tool instance adds one schema; visible delegation tools share one short system-prompt section. +Fixed schema cost per parent request; `modelSelectionSettings` adds three optional route fields. Each tool instance adds one schema; visible delegation tools share one short system-prompt section. #### KV Cache effect @@ -143,11 +142,11 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -The shared `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. +The `list_subagent_models` definition registered by a settings-backed instance reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. #### Token effect -One fixed discovery schema is present in shipped compositions, regardless of the model-selection preference. Route metadata enters the transcript only when the model calls the tool. +One fixed discovery schema is present in compositions with `modelSelectionSettings`, regardless of the Session’s preference; the shipped Web presets include it, while headless and SDK profiles do not. Route metadata enters the transcript only when the model calls the tool. #### KV Cache effect @@ -209,8 +208,8 @@ Append-only; newly visible content follows the reusable request prefix and does These limits define what this tool does not return or enforce; they are current package constraints. - **This tool does not wait for results** — use the completion notice or child messages. Programmatic result waiting belongs to the subagent service. -- **Fork tools reject model selection** — tools that inherit parent context neither expose nor accept `provider`, `model`, or `reasoning_effort`, even with Session authorization. Configured `agentOptions` remain available to the deployment. -- **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection is restricted to the Session’s authorized routes, or the current parent route without an enabled policy; unsupported backends reject selection fields. +- **Fork tools reject model selection** — tools that inherit parent context neither expose nor accept `provider`, `model`, or `reasoning_effort`, even with Session authorization, and `modelSelectionSettings` fails their mount. Configured `agentOptions` remain available to the deployment. +- **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection requires `modelSelectionSettings` and is restricted to the Session’s authorized routes, or the current parent route without an enabled policy. ### Dev Note diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index 5e97c8db7a..7a819691ca 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -36,7 +36,6 @@ kind: "package-reference" ```yaml - name: '@deepseek-ai/dsh-subagent' - name: '@deepseek-ai/dsh-subagent-spawn-in-process' -- name: '@deepseek-ai/dsh-tool-subagent/list-models' - name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn @@ -47,7 +46,7 @@ kind: "package-reference" |---|---|---| | `provider` | 必填 | `ctx.subagents` 上的提供方名称（如 `spawn`、`fork`、`acp`） | | `toolName` | `subagent` | 面向模型的工具名称；每个已加载实例必须不同 | -| `modelSelectionSettings` | `false` | 为每个顶层 Session 读取宿主的精确路由授权偏好；常驻 preset 观察匹配 Session，直接 Agent setup 则显式传入其 Session；要求提供方支持 `agentOptions` | +| `modelSelectionSettings` | `false` | 公开子级路由字段和 `list_subagent_models`，并为每个顶层 Session 读取宿主的精确路由授权偏好；常驻 preset 观察匹配 Session，直接 Agent setup 则显式传入其 Session；要求提供方支持 `agentOptions` 且不继承父级上下文 | | `agentOptions` | — | 配置的子级 `provider`、`model`、适配器所有的 `reasoningEffort` 与正整数 `maxTokens` 默认值；要求提供方支持 `agentOptions`，并会覆盖提供方持有的路由默认值 | | `persona` | — | 每个子 agent 独立的 persona；要求提供方具备 `persona` 能力 | | `toolFilter` | — | 每个子 agent 独立的全局工具限制；要求提供方具备 `toolFilter` 能力 | @@ -65,7 +64,7 @@ kind: "package-reference" ### 选择子级 LLM -支持 `agentOptions` 且不继承父级上下文的后端始终公开可选的 `provider`、`model` 和 `reasoning_effort` 字段。在委派工具旁加载一次 `@deepseek-ai/dsh-tool-subagent/list-models` 即可共享发现功能；随附组合已包含它。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。设置 `modelSelectionSettings: true` 后，新顶层 Session 会读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有策略的恢复 Session 仍受父级路由限制。ACP、Codex 和 Claude Code 不支持 `agentOptions`，会拒绝选择字段。 +设置 `modelSelectionSettings: true` 后，该实例公开可选的 `provider`、`model` 和 `reasoning_effort` 字段，并注册 `list_subagent_models`；未设置的实例两者都不公开，并拒绝选择字段。该实例为新顶层 Session 读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。此模式要求后端公布 `agentOptions` 且不继承父级对话；进程内 Spawn 和 DSH SDK 满足条件，Fork、ACP、Codex 和 Claude Code 会使挂载失败。 一次调用需同时提供 `provider` 与 `model`；当配置值、父 agent 值或提供方持有的默认值能提供路由时，也可只提供推理等级。静态的 `provider.agentRouteDefaults` 在存在时构成提供方／模型基线；工具配置与模型字段会在路由相关强度合并和确切路由预检前覆盖它。没有这些默认值的提供方会使用父 agent 最新已记录请求中的兼容值，再使用父级首次请求前的创建选项，并保留配置的 `maxTokens`。更改路由但未显式提供推理等级时，会清除继承的路由自有等级，使所选模型解析自己的默认值。实时 LLM 适配器在创建子 agent 前校验有效路由。目录成员资格只提供建议，因此适配器接受时，模型可以使用未列出的 id。 @@ -81,7 +80,7 @@ kind: "package-reference" ### 设计理念 -一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。读取设置的实例在 Agent 创建时记录 Session 授权；调用时独立于注册读取持久策略。直接配置 Agent 时显式传入尚未发布的 Session。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。发现功能由一个独立组合项提供。本包不导出运行时不变量：策略直接来自 Session 投影，并在执行时校验，没有另行保存的逐 Agent 定义状态需要协调。 +一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。读取设置的实例还会在同一组合中注册 `list_subagent_models`，并在 Agent 创建时记录 Session 授权；调用时独立于注册读取持久策略。直接配置 Agent 时显式传入尚未发布的 Session。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。由于 `list_subagent_models` 使用全局名称，一个工具作用域中最多只有一个实例可以设置 `modelSelectionSettings`。本包不导出运行时不变量：策略直接来自 Session 投影，并在执行时校验。 ### Activation 所有权 @@ -129,11 +128,11 @@ kind: "package-reference" #### 模型看到什么 -生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。支持 `agentOptions` 且不继承父级上下文的后端始终公开 `provider`、`model` 和 `reasoning_effort`。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 +生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。设置了 `modelSelectionSettings` 的实例在每个 Session 中都会增加 `provider`、`model`、`reasoning_effort` 及路由指导。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 #### Token 影响 -每次父级请求承担固定 schema 成本，其中支持选择的后端包含三个可选路由字段。每个工具实例增加一份 schema；可见委派工具共享一个简短系统提示词段。 +每次父级请求承担固定 schema 成本；`modelSelectionSettings` 会增加三个可选路由字段。每个工具实例增加一份 schema；可见委派工具共享一个简短系统提示词段。 #### KV Cache 影响 @@ -143,11 +142,11 @@ kind: "package-reference" #### 模型看到什么 -共享的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 +读取设置的实例注册的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 #### Token 影响 -随附组合中始终存在一份固定发现 schema，不受模型选择偏好影响。只有模型调用工具时，路由元数据才进入 transcript。 +设置了 `modelSelectionSettings` 的组合中存在一份固定发现 schema，不受 Session 偏好影响；随附的 Web preset 包含它，headless 与 SDK profile 不包含。只有模型调用工具时，路由元数据才进入 transcript。 #### KV Cache 影响 @@ -209,8 +208,8 @@ Start independent delegations with `subagent` or `subagent_fork` together in one 这些限制说明本工具不返回或不强制执行什么；它们是当前包约束。 - **本工具不等待结果**——使用完成通知或子任务消息。程序化结果等待由 subagent 服务提供。 -- **fork 工具拒绝模型选择** — 继承父级上下文的工具既不公开也不接受 `provider`、`model` 或 `reasoning_effort`，即使 Session 已授权也不例外。部署仍可配置 `agentOptions`。 -- **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择受 Session 授权路由限制，没有启用策略时只允许当前父级路由；不支持选择的后端会拒绝选择字段。 +- **fork 工具拒绝模型选择** — 继承父级上下文的工具既不公开也不接受 `provider`、`model` 或 `reasoning_effort`，即使 Session 已授权也不例外；`modelSelectionSettings` 会使其挂载失败。部署仍可配置 `agentOptions`。 +- **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择需要 `modelSelectionSettings`，并受 Session 授权路由限制，没有启用策略时只允许当前父级路由。 ### 开发备注 diff --git a/packages/subagent/tool-subagent/package.json b/packages/subagent/tool-subagent/package.json index fbc8d717d5..a535e39446 100644 --- a/packages/subagent/tool-subagent/package.json +++ b/packages/subagent/tool-subagent/package.json @@ -24,10 +24,6 @@ }, "./src/*": "./src/*", "./package.json": "./package.json", - "./list-models": { - "types": "./lib/types/list-models.d.ts", - "default": "./lib/list-models.js" - }, "./locale/*.json": "./locale/*.json", "./model-selection-settings/locale/*.json": "./locale/model-selection-settings/*.json" }, @@ -35,7 +31,6 @@ "locale/*.json", "locale/model-selection-settings/*.json", "lib/index.js", - "lib/list-models.js", "lib/model-selection-settings.js", "lib/types/**/*.d.ts" ], diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index 19af35e38f..928e405f9f 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -29,6 +29,7 @@ import { requestedAgentOptions, } from './model-selection.ts' import type { DelegationModelRequest } from './model-selection.ts' +import { registerListSubagentModels } from './list-models.ts' import type {} from './model-selection-settings.ts' import { recordSubagentModelSelection, @@ -53,8 +54,11 @@ export interface Config { */ toolName?: string /** - * Sample the Host `subagent-model-selection` setting for each new top-level - * Session and inherit that decision in its child Sessions. + * Expose optional child `provider`, `model`, and `reasoning_effort` fields + * plus `list_subagent_models`, and sample the Host `subagent-model-selection` + * setting for each new top-level Session; child Sessions inherit that + * decision. Requires a provider with `agentOptions` that does not inherit + * the parent conversation. At most one instance per tool scope may enable it. */ modelSelectionSettings?: boolean /** @@ -169,7 +173,7 @@ export function apply(ctx: Context, config: Config, session?: Session): void { } const toolName = config.toolName ?? 'subagent' - const modelSelectionCapable = config.modelSelectionSettings === true + const modelSelectionEnabled = config.modelSelectionSettings === true ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) const assertSubagentProviderConfiguration = (subagentProvider: SubagentProvider): void => { @@ -184,12 +188,17 @@ export function apply(ctx: Context, config: Config, session?: Session): void { `tool-subagent: provider "${subagentProvider.name}" does not support child agentOptions`, ) } - if (modelSelectionCapable && !subagentProvider.capabilities.agentOptions) { + if (modelSelectionEnabled && !subagentProvider.capabilities.agentOptions) { throw new Error( `tool-subagent: provider "${subagentProvider.name}" does not support child model selection`, ) } - + if (modelSelectionEnabled && subagentProvider.inheritsParentContext) { + throw new Error( + `tool-subagent: provider "${subagentProvider.name}" inherits the parent conversation and does not accept child model selection` + + ' — remove `modelSelectionSettings` from this instance', + ) + } } // Validate provider-owned config outside the optional LLM binding so an @@ -201,12 +210,18 @@ export function apply(ctx: Context, config: Config, session?: Session): void { if (initialProvider !== undefined) assertSubagentProviderConfiguration(initialProvider) const install = (runtimeCtx: Context): void => { + if (modelSelectionEnabled) { + registerListSubagentModels(runtimeCtx, (exec) => { + if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') + const routes = subagentModelSelectionPolicy(runtimeCtx.sessionProjections, exec.agent.session) + return routes === undefined ? undefined : { routes } + }) + } // Load order and HMR replacement can change provider availability while // this fiber remains active. let mounted: { subagentProvider: SubagentProvider; disposeTool: () => void } | undefined const mount = (subagentProvider: SubagentProvider): void => { assertSubagentProviderConfiguration(subagentProvider) - const modelSelectionEnabled = subagentProvider.capabilities.agentOptions && !subagentProvider.inheritsParentContext const wording = providerWording(subagentProvider.inheritsParentContext) const providerRouteDefaults = subagentProvider.agentRouteDefaults const selectionDescription = providerRouteDefaults !== undefined diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index 5c100eedf6..a25c9db8c8 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -6,7 +6,6 @@ import type { LlmProviderInfo } from '@deepseek-ai/dsh-llm' import { defineTool } from '@deepseek-ai/dsh-tools' import { parentAgentOptionsForDelegation } from '@deepseek-ai/dsh-subagent' import type { ToolRunContext } from '@deepseek-ai/dsh-tools' -import { subagentModelSelectionPolicy, subagentModelSelectionProjectionDefinition } from './model-selection-state.ts' import type { ModelSelectionPolicy } from './model-selection.ts' interface ListSubagentModelsRequest { @@ -95,8 +94,8 @@ async function listSubagentModels( } /** - * Register `list_subagent_models` in the owning composition. - * Provider listings intersect the catalog with authorization, or resolve only the parent model when selection is disabled. + * Register `list_subagent_models` for the delegation-tool instance that owns model selection. + * Provider listings intersect the catalog with authorization, or resolve only the parent model without a Session policy. * @param ctx - Context whose tool registry owns the fixed discovery definition. * @param policyFor - Read the calling Session’s authorized routes; undefined restricts discovery to its parent model. */ @@ -126,21 +125,3 @@ export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolR }, })) } - -/** Shared discovery plugin for all delegation tools in one composition. */ -export const name = 'tool-subagent-list-models' -/** Services used to register discovery and read durable route policy. */ -export const inject = ['tools', 'sessionProjections'] - -/** - * Register one discovery tool whose results follow the calling Session’s policy. - * @param ctx - Owning composition context. - */ -export function apply(ctx: Context): void { - ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) - registerListSubagentModels(ctx, (exec) => { - if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') - const routes = subagentModelSelectionPolicy(ctx.sessionProjections, exec.agent.session) - return routes === undefined ? undefined : { routes } - }) -} diff --git a/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs b/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs deleted file mode 100644 index 871ddc0ffc..0000000000 --- a/packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs +++ /dev/null @@ -1,38 +0,0 @@ -/** Deterministic metadata lookups for the model-discovery transcript. */ -import { LlmAdapter } from '@deepseek-ai/dsh-llm' - -class DiscoveryAdapter extends LlmAdapter { - async listModels(provider) { - return [ - { provider, id: 'healthy', name: 'Healthy' }, - { provider, id: 'unauthorized', name: 'Unauthorized' }, - ] - } - - async resolveModel(provider, model) { - if (model === 'removed') throw new Error('model was removed') - return { provider, id: model, name: 'Healthy' } - } - - async *stream() { - throw new Error('metadata-only fixture does not serve model requests') - } -} - -export const name = 'discovery-fixture' -export const inject = ['llm'] - -export function apply(ctx) { - ctx.llm.registerAdapter(['discovery-fixture'], new DiscoveryAdapter()) - ctx.on('agent/pre-step', async ({ agent, turn, step }, next) => { - if (turn === 1 && step === 1) { - agent.session.append('subagent/model-selection-policy', { - allowedModels: [ - { provider: 'discovery-fixture', model: 'removed' }, - { provider: 'discovery-fixture', model: 'healthy' }, - ], - }) - } - return next() - }) -} diff --git a/packages/subagent/tool-subagent/tests/harness.ts b/packages/subagent/tool-subagent/tests/harness.ts index 061161673e..01279f3b13 100644 --- a/packages/subagent/tool-subagent/tests/harness.ts +++ b/packages/subagent/tool-subagent/tests/harness.ts @@ -9,7 +9,6 @@ import SubagentRuntime from '@deepseek-ai/dsh-subagent' import { SessionId } from '@deepseek-ai/dsh-session' import * as mock from './scripted-provider.ts' import * as tool from '../src/index.ts' -import * as discovery from '../src/list-models.ts' import SubagentModelSelectionConfig from '../src/model-selection-settings.ts' const contexts = new Set() @@ -35,6 +34,8 @@ let setupAgentCounter = 0 /** Test-only opt-in translated to the real Host setting and Session path. */ type SetupConfig = tool.Config & { withModelSelection?: boolean + /** Host setting sampled by a `withModelSelection` setup; defaults to enabled. */ + hostSelectionEnabled?: boolean parentAgentOptions?: AgentOptions } @@ -49,17 +50,16 @@ const TEST_ALLOWED_MODELS = [ export async function setup(toolConfig: SetupConfig, mockConfig: Partial = {}): Promise { const ctx = ownContext(new Context()) - const { withModelSelection, parentAgentOptions, ...config } = toolConfig + const { withModelSelection, hostSelectionEnabled = true, parentAgentOptions, ...config } = toolConfig if (withModelSelection === true) { await ctx.plugin(SubagentModelSelectionConfig, { - enabled: true, + enabled: hostSelectionEnabled, allowedModels: TEST_ALLOWED_MODELS, }) await mountAgentLoopTestDependencies(ctx) await ctx.plugin(AgentLoop, { agents: [] }) await mountWorkingDirectoryFixture(ctx) await ctx.plugin(SubagentRuntime) - await ctx.plugin(discovery) const provider = await mock.mountScriptedProvider(ctx, { name: 'mock', ...mockConfig }) setupProviders.set(ctx, provider) const handle = await ctx.agents.create({ @@ -82,7 +82,6 @@ export async function setup(toolConfig: SetupConfig, mockConfig: Partial { - it('is loaded through its own composition entry', async () => { + it('is omitted unless its delegation-tool instance owns model selection', async () => { const ctx = new Context() await ctx.plugin(SessionProjectionRegistry) await ctx.plugin(LlmRuntime) @@ -114,14 +116,23 @@ describe('list_subagent_models', () => { it('requires a calling Agent and an effective parent route', async () => { const ctx = new Context() + await ctx.plugin(SubagentModelSelectionConfig) await ctx.plugin(SessionProjectionRegistry) await ctx.plugin(SystemPrompt) await ctx.plugin(ToolRuntime) - await ctx.plugin({ apply, inject, name }) + await mountWorkingDirectoryFixture(ctx) + await ctx.plugin(SubagentRuntime) + await ctx.plugin({ + name: 'direct-tool-subagent', + inject: tool.inject, + apply: (direct: Context) => { + tool.apply(direct, { provider: 'unused', modelSelectionSettings: true }, Session.create(SessionId('direct-discovery'))) + }, + }) expect(text(await call(ctx, {}))).toContain('requires a calling agent') await ctx.fiber.dispose() - const parentContext = await setup({ provider: 'mock' }) + const parentContext = await setup({ provider: 'mock', withModelSelection: true, hostSelectionEnabled: false }) const agent = modelSelectionSetupAgent(parentContext) const result = await parentContext.tools.execute({ signal: testToolSignal, callId: ToolCallId('missing-parent-route'), @@ -131,7 +142,7 @@ describe('list_subagent_models', () => { expect(text(result)).toContain('without a parent provider and model') }) - it('reads the recorded Session allowlist through the shared discovery definition', async () => { + it('reads the recorded Session allowlist through the settings-owned discovery definition', async () => { const ctx = await setup({ provider: 'mock', withModelSelection: true }) ctx.llm.registerAdapter(['alpha', 'secret'], new CatalogAdapter()) const result = await ctx.tools.execute({ @@ -279,8 +290,13 @@ describe('list_subagent_models', () => { } }) - it.each([false, true])('resolves only the current parent model when selection is disabled (empty catalog: %s)', async (empty) => { - const ctx = await setup({ provider: 'mock', parentAgentOptions: { provider: 'alpha', model: 'unlisted' } }) + it.each([false, true])('resolves only the current parent model without a Session policy (empty catalog: %s)', async (empty) => { + const ctx = await setup({ + provider: 'mock', + withModelSelection: true, + hostSelectionEnabled: false, + parentAgentOptions: { provider: 'alpha', model: 'unlisted' }, + }) const adapter = new CatalogAdapter(empty) const listModels = vi.spyOn(adapter, 'listModels') ctx.llm.registerAdapter(['alpha'], adapter) diff --git a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts index a4dc33b0fa..ff374ed096 100644 --- a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts @@ -14,7 +14,6 @@ import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection' import SubagentRuntime from '@deepseek-ai/dsh-subagent' import * as SubagentSpawn from '@deepseek-ai/dsh-subagent-spawn-in-process' import * as tool from '../src/index.ts' -import * as discovery from '../src/list-models.ts' import SubagentModelSelectionConfig from '../src/model-selection-settings.ts' import { subagentModelSelectionPolicy, @@ -49,10 +48,8 @@ async function boot(withPreset = true): Promise { await mountWorkingDirectoryFixture(ctx) subagentConfigs.set(ctx, await liveConfig(ctx, SubagentRuntime)) await ctx.plugin(SubagentSpawn, { providerName: 'spawn' }) - await ctx.plugin(discovery) if (withPreset) { const preset = createScope(ctx, { preset: 'model-selection-test' }) - await preset.ctx.plugin(discovery) await preset.ctx.plugin(tool, { provider: 'spawn', modelSelectionSettings: true, @@ -272,7 +269,6 @@ describe('SubagentModelSelectionConfig', () => { const ctx = await boot(false) const preset = createScope(ctx, { preset: 'standard' }) const other = createScope(ctx, { preset: 'minimal' }) - await preset.ctx.plugin(discovery) await preset.ctx.plugin(tool, { provider: 'spawn', modelSelectionSettings: true, diff --git a/packages/subagent/tool-subagent/tests/model-selection.spec.ts b/packages/subagent/tool-subagent/tests/model-selection.spec.ts index 4a99144c81..69b72ae876 100644 --- a/packages/subagent/tool-subagent/tests/model-selection.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection.spec.ts @@ -141,13 +141,37 @@ describe('dsh-tool-subagent model selection', () => { expect(text(result)).toContain('model selection is disabled for this tool instance') }) - it('keeps route fields visible but rejects other routes when selection is disabled', async () => { + it('hides and rejects route fields without modelSelectionSettings', async () => { const ctx = await setup({ provider: 'mock', parentAgentOptions: { provider: 'alpha', model: 'parent-model' } }) const schema = ctx.tools.schemas().find(entry => entry.name === 'subagent')! const props = (schema.parameters as { properties?: Record }).properties ?? {} + expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'prompt']) + expect(schema.description).not.toContain('list_subagent_models') + expect(ctx.tools.get('list_subagent_models')).toBeUndefined() + + for (const selection of [ + { provider: 'alpha', model: 'parent-model' }, + { reasoning_effort: 'low' }, + ]) { + const result = await callSubagent(ctx, { description: 'forced route', prompt: 'do it', ...selection }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('child model selection is disabled for this tool instance') + } + }) + + it('keeps settings-owned route fields visible but rejects other routes without a Session policy', async () => { + const ctx = await setup({ + provider: 'mock', + withModelSelection: true, + hostSelectionEnabled: false, + parentAgentOptions: { provider: 'alpha', model: 'parent-model' }, + }) + const agent = modelSelectionSetupAgent(ctx) + const schema = ctx.tools.schemas(agent).find(entry => entry.name === 'subagent')! + const props = (schema.parameters as { properties?: Record }).properties ?? {} expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'model', 'prompt', 'provider', 'reasoning_effort']) expect(schema.description).toContain('list_subagent_models') - expect(ctx.tools.get('list_subagent_models')).toBeDefined() + expect(ctx.tools.get('list_subagent_models', agent)).toBeDefined() const result = await callSubagent(ctx, { description: 'forced route', @@ -159,10 +183,12 @@ describe('dsh-tool-subagent model selection', () => { expect(text(result)).toContain('is not allowed for this Session') }) - it('discovers only the parent route by default and permits its explicit route and effort', async () => { + it('discovers only the parent route without a Session policy and permits its explicit route and effort', async () => { const requests: SubagentStartRequest[] = [] const ctx = await setup({ provider: 'mock', + withModelSelection: true, + hostSelectionEnabled: false, parentAgentOptions: parentWithRoute().options, }, { onStart: (request) => { requests.push(request) } }) ctx.llm.registerAdapter(['alpha', 'other-provider'], new MockAdapter([], REASONING)) @@ -469,17 +495,27 @@ describe('dsh-tool-subagent model selection', () => { expect(resolve).not.toHaveBeenCalled() }) + it('rejects modelSelectionSettings on a provider that inherits the parent conversation', async () => { + await expect(setup({ provider: 'mock', withModelSelection: true }, { inheritsParentContext: true })) + .rejects.toThrow('provider "mock" inherits the parent conversation and does not accept child model selection') + }) + it('keeps fork model selection unavailable even with Session authorization', async () => { const started = vi.fn() - const ctx = await setup({ provider: 'mock', withModelSelection: true }, { + const ctx = await setup({ provider: 'mock' }, { inheritsParentContext: true, onStart: started, }) - const schema = ctx.tools.schemas(modelSelectionSetupAgent(ctx)).find(entry => entry.name === 'subagent')! + const agent = modelSelectionSetupAgent(ctx) + agent.session.append('subagent/model-selection-policy', { + allowedModels: [{ provider: 'alpha', model: 'fast-model' }], + }) + const schema = ctx.tools.schemas(agent).find(entry => entry.name === 'subagent')! const props = (schema.parameters as { properties?: Record }).properties ?? {} expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'prompt']) expect(schema.description).toContain('inherits this conversation') - expect(schema.description).not.toContain('Child LLM selection') + expect(schema.description).not.toContain('list_subagent_models') + expect(ctx.tools.get('list_subagent_models', agent)).toBeUndefined() for (const selection of [ { provider: 'alpha', model: 'fast-model' }, diff --git a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts index 4a0a1a6a08..fb39038860 100644 --- a/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts +++ b/packages/subagent/tool-subagent/tests/tool-subagent.spec.ts @@ -102,10 +102,7 @@ describe('dsh-tool-subagent', () => { expect(Object.keys(props).sort()).toEqual([ 'cwd', 'description', - 'model', 'prompt', - 'provider', - 'reasoning_effort', ]) expect(schema!.description).not.toContain('job_output') }) diff --git a/packages/subagent/tool-subagent/tsdown.config.ts b/packages/subagent/tool-subagent/tsdown.config.ts index 8cdb8ec132..4a467f2c5e 100644 --- a/packages/subagent/tool-subagent/tsdown.config.ts +++ b/packages/subagent/tool-subagent/tsdown.config.ts @@ -15,5 +15,4 @@ const entry = (path: string) => ({ export default defineConfig([ entry('lib/types/index.js'), entry('lib/types/model-selection-settings.js'), - entry('lib/types/list-models.js'), ]) diff --git a/scripts/check-workspace-constraints.ts b/scripts/check-workspace-constraints.ts index 7ea4d2e0aa..1880c55d1e 100644 --- a/scripts/check-workspace-constraints.ts +++ b/scripts/check-workspace-constraints.ts @@ -195,8 +195,9 @@ const packageFileExtras: Readonly> = { '@deepseek-ai/dsh-experimental-inspector': ['lib/client.*.js', 'lib/worker.js', 'lib/devtools/**'], // Creator's composition guidance travels with the declaration package. '@deepseek-ai/dsh-agent-preset': ['skills'], - // Settings and shared discovery have independent composition entries. - '@deepseek-ai/dsh-tool-subagent': ['lib/list-models.js', 'lib/model-selection-settings.js'], + // The Web Host mounts the default-off settings owner independently of the + // delegation-tool instances. + '@deepseek-ai/dsh-tool-subagent': ['lib/model-selection-settings.js'], // The JSONL backend resolves its private verification Worker relative to // import.meta.url; it is shipped without a public package subpath. '@deepseek-ai/dsh-session-persistence-jsonl': ['lib/worker.cjs'], diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index 81b10491e0..4cfa955c2d 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -73,7 +73,7 @@ import * as PluginManagerTools from '@deepseek-ai/dsh-plugin-manager/tools' import SandboxPolicy from '@deepseek-ai/dsh-sandbox-policy' import McpResources from '@deepseek-ai/dsh-mcp-resources' import * as ToolSubagent from '@deepseek-ai/dsh-tool-subagent' -import * as SubagentModelDiscovery from '../packages/subagent/tool-subagent/src/list-models.ts' +import { registerListSubagentModels } from '../packages/subagent/tool-subagent/src/list-models.ts' import * as ToolWeb from '@deepseek-ai/dsh-tool-web' import WorkflowEngine from '@deepseek-ai/dsh-workflow' import type { WorkflowRun, WorkflowStartRequest } from '@deepseek-ai/dsh-workflow' @@ -519,10 +519,10 @@ const TOOL_PACKAGES: ToolPackage[] = [ await ctx.plugin(LlmRuntime) registerCatalogSubagentProvider(ctx, 'mock') await ctx.plugin(ToolSubagent, { provider: 'mock' }) - await ctx.plugin(SubagentModelDiscovery) + registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'mock', model: 'mock' }] })) }, note: - 'The registered delegation name is the load-time `toolName` config (default `subagent`); backends supporting agentOptions without inheriting parent context always expose optional route fields, and the shared discovery plugin resolves the calling Session’s allowed routes. Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions; `subagent_fork` rejects model-selection arguments; without an enabled policy, spawn selection and discovery are restricted to the current parent route. Each instance selects its backend through `provider` and reads model-selection settings through `modelSelectionSettings`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', + 'The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', }, { pkg: '@deepseek-ai/dsh-tool-subagent-control', diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/result.json b/scripts/snapshots/python-sdk-single-exe/advanced/result.json index 333c0cd655..0cb043baf0 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/result.json +++ b/scripts/snapshots/python-sdk-single-exe/advanced/result.json @@ -207,7 +207,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -304,7 +303,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -1763,7 +1761,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -1890,7 +1887,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -2961,7 +2957,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "snapshot_double", "subagent", "workflow", @@ -3720,7 +3715,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "snapshot_double", "subagent", "workflow", diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl index 2c5e9c5f05..a169f4f81a 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/advanced/session.1.v4.jsonl @@ -13,7 +13,7 @@ {"type":"user/message","data":{"id":"{{messageId}}","role":"user","content":[{"type":"text","text":"Serial agent creation completed."}],"source":{"kind":"plugin:serial-created-fixture"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Reply with exactly DIRECT_CHILD_OK and nothing else."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).\n\nYou are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."},{"name":"subagent:delegation","text":"You are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Reply with exactly DIRECT_CHILD_OK and","messageSeqs":[12],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{child-1}}","sessionFormatVersion":4,"throughSeq":16}} diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl index 5ec93b7783..88aaf9d5e2 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/advanced/session.2.v4.jsonl @@ -13,7 +13,7 @@ {"type":"user/message","data":{"id":"{{messageId}}","role":"user","content":[{"type":"text","text":"Serial agent creation completed."}],"source":{"kind":"plugin:serial-created-fixture"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Reply with exactly WORKFLOW_CHILD_OK and nothing else."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).\n\nYou are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."},{"name":"subagent:delegation","text":"You are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","reasoningEffort":"high","maxTokens":256000},"adapterDefaults":{"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Reply with exactly WORKFLOW_CHILD_OK and","messageSeqs":[12],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{child-2}}","sessionFormatVersion":4,"throughSeq":16}} diff --git a/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl index ed4f842bde..dcd09c277c 100644 --- a/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/advanced/session.v4.jsonl @@ -13,12 +13,12 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Run the advanced packaged-runtime snapshot scenario."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"id":"{{messageId}}","role":"user","content":[{"type":"image","attachment":{"attachmentId":"sha256:b1ff9c8ea3a780bad09b346c423d2d0e46815926879b18e841d928376a946640","mediaType":"image/png","width":1,"height":1,"bytes":69}}],"source":{"kind":"python-snapshot-image-offload"}},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Run the advanced packaged-runtime snapsh","messageSeqs":[11],"source":{"kind":"fallback"}}} {"type":"assistant/attempt","data":{"turn":1,"step":1,"stream":[{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"error","failure":{"code":"IMAGE_OFFLOAD_REQUIRED","message":"authored image budget failure","offloadImages":1}}}}]}} {"type":"image/offload","data":{"targets":[{"seq":13,"imageIndexes":[0]}]}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","snapshot_double","subagent","workflow","working_directory"]},"reason":"series"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","snapshot_double","subagent","workflow","working_directory"]},"reason":"series"}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{parent}}","sessionFormatVersion":4,"throughSeq":19}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-native","name":"snapshot_double","arguments":"{\"value\": 21}"}],"source":{"kind":"model","provider":"deepseek-official","model":"smoke-model","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"}]}},"id":"{{messageId}}"},"usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"advanced-native","name":"snapshot_double","args":[""]},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"advanced-native","args":["{\"value\": 21}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-native","name":"snapshot_double","arguments":"{\"value\": 21}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"}]}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"advanced-native","name":"snapshot_double","arguments":"{\"value\": 21}"}} diff --git a/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json b/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json index 8637546c53..d93f0c33c1 100644 --- a/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json +++ b/scripts/snapshots/python-sdk-single-exe/dynamic-tools/tool-history.json @@ -56,7 +56,6 @@ "job_list", "job_output", "list_agents", - "list_subagent_models", "read", "read_image", "send_message", @@ -113,7 +112,6 @@ "job_list", "job_output", "list_agents", - "list_subagent_models", "read", "read_image", "send_message", @@ -185,7 +183,6 @@ "job_list", "job_output", "list_agents", - "list_subagent_models", "read", "read_image", "send_message", diff --git a/scripts/snapshots/python-sdk-single-exe/restart/requests.json b/scripts/snapshots/python-sdk-single-exe/restart/requests.json index 3c4c8b10d3..06bbf97d4f 100644 --- a/scripts/snapshots/python-sdk-single-exe/restart/requests.json +++ b/scripts/snapshots/python-sdk-single-exe/restart/requests.json @@ -23,7 +23,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "subagent", "workflow", "working_directory" @@ -53,7 +52,6 @@ "job_kill", "job_list", "job_output", - "list_subagent_models", "subagent", "workflow", "working_directory" diff --git a/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl index 3ce01a7a4d..59d1880f27 100644 --- a/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/restart/session.1.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{messageId}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Complete the first isolated Python SDK process turn."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Complete the first isolated Python","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{session-1}}","sessionFormatVersion":4,"throughSeq":12}} diff --git a/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl index cc12894292..be7dd18e39 100644 --- a/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/restart/session.2.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{messageId}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Complete the second isolated Python SDK process turn."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","subagent","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","subagent","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Complete the second isolated Python","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"session-log-deepseek/delivery-accepted","data":{"sessionId":"{{session-2}}","sessionFormatVersion":4,"throughSeq":12}} diff --git a/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl b/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl index fc8d44ed52..c679ff36b7 100644 --- a/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl +++ b/scripts/snapshots/python-sdk-single-exe/scheduler-recovery/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{messageId}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Run three todo updates; the scheduler will reject the second call."}],"source":{"kind":"user"},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{messageId}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","list_subagent_models","subagent","todo_write","workflow","working_directory"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"smoke-model","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["cordis_inspect_list","cordis_inspect_query","job_kill","job_list","job_output","subagent","todo_write","workflow","working_directory"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"smoke-model","contextWindow":1000000}} {"type":"session/title","data":{"title":"Run three todo updates; the","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"scheduler-complete","name":"todo_write","arguments":"{\"todos\": []}"},{"type":"tool-call","id":"scheduler-fail","name":"todo_write","arguments":"{\"todos\": []}"},{"type":"tool-call","id":"scheduler-unstarted","name":"todo_write","arguments":"{\"todos\": []}"}],"source":{"kind":"model","provider":"deepseek-official","model":"smoke-model","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"},{"type":"tool-call"},{"type":"tool-call"}]}},"id":"{{messageId}}"},"usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"scheduler-complete","name":"todo_write","args":[""]},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"scheduler-complete","args":["{\"todos\": []}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"scheduler-complete","name":"todo_write","arguments":"{\"todos\": []}"}}},{"type":"chunk","time":0,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":1,"dt":[],"id":"scheduler-fail","name":"todo_write","args":[""]},{"type":"tool-call-chunks","time0":0,"index":1,"dt":[],"id":"scheduler-fail","args":["{\"todos\": []}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"scheduler-fail","name":"todo_write","arguments":"{\"todos\": []}"}}},{"type":"chunk","time":0,"chunk":{"type":"block-start","index":2,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":2,"dt":[],"id":"scheduler-unstarted","name":"todo_write","args":[""]},{"type":"tool-call-chunks","time0":0,"index":2,"dt":[],"id":"scheduler-unstarted","args":["{\"todos\": []}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":2,"block":{"type":"tool-call","id":"scheduler-unstarted","name":"todo_write","arguments":"{\"todos\": []}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3,"totalTokens":6}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"smoke-model"},"blocks":[{"type":"tool-call"},{"type":"tool-call"},{"type":"tool-call"}]}}}]},"surfaceOp":"append"} diff --git a/snapshots/sdk/bash-tool/tool-schemas.expected.json b/snapshots/sdk/bash-tool/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/bash-tool/tool-schemas.expected.json +++ b/snapshots/sdk/bash-tool/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl b/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl index b497f77dbd..bfec6f1e36 100644 --- a/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl +++ b/snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl @@ -6,7 +6,7 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"system/message","seq":7,"time":0,"data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":8,"time":0,"data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":9,"time":0,"data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/context","seq":11,"time":0,"data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"session/title","seq":12,"time":0,"data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":13,"time":0,"data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} @@ -14,14 +14,14 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/result","seq":15,"time":0,"data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_task_read"},"toolCallId":"call_task_read","content":[{"type":"text","text":"{{cwd}}/task.txt\nfile\n\n1: Reply with the single word DONE after reading this file.\n\n(End of file - total 1 lines)\n"}],"isError":false,"id":"{{sessionId}}"},"meta":{"path":"{{cwd}}/task.txt","offset":1,"lines":[{"number":1,"text":"Reply with the single word DONE after reading this file."}],"totalLines":1}},"sourceEventSeqs":[14],"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/end","seq":16,"time":0,"data":{"turn":1,"step":1}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/start","seq":17,"time":0,"data":{"turn":1,"step":2}}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":18,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":18,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":19,"time":0,"data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":18},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":20,"time":0,"data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/call","seq":21,"time":0,"data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/result","seq":22,"time":0,"data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{sessionId}}"}},"sourceEventSeqs":[21],"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/end","seq":23,"time":0,"data":{"turn":1,"step":2}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/start","seq":24,"time":0,"data":{"turn":1,"step":3}}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":25,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":25,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":26,"time":0,"data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{sessionId}}"}},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":27,"time":0,"data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":0,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":0,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/end","seq":28,"time":0,"data":{"turn":1,"step":3}}}} diff --git a/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl b/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl index 8472750d16..aee79405da 100644 --- a/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl +++ b/snapshots/sdk/dynamic-tool-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790049391235,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391235,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790049391235,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -17,14 +17,14 @@ {"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_task_read"},"toolCallId":"call_task_read","content":[{"type":"text","text":"{{cwd}}/task.txt\nfile\n\n1: Reply with the single word DONE after reading this file.\n\n(End of file - total 1 lines)\n"}],"isError":false,"id":"{{message:5}}"},"meta":{"path":"{{cwd}}/task.txt","offset":1,"lines":[{"number":1,"text":"Reply with the single word DONE after reading this file."}],"totalLines":1}},"sourceEventSeqs":[14],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:6}}"},"headerSeq":18},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:7}}"},"stream":[{"type":"chunk","time":1790049391255,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391255,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790049391255,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790049391255,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:8}}"}},"sourceEventSeqs":[21],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:9}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:10}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790049391270,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790049391270,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790049391270,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/sdk/persistent-tools/cordis.yml b/snapshots/sdk/persistent-tools/cordis.yml index e2ee87e8a7..6262d2fb44 100644 --- a/snapshots/sdk/persistent-tools/cordis.yml +++ b/snapshots/sdk/persistent-tools/cordis.yml @@ -84,10 +84,6 @@ name: '@deepseek-ai/dsh-tool-subagent-control' disabled: true -- id: tool-subagent-list-models - name: '@deepseek-ai/dsh-tool-subagent/list-models' - disabled: true - - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' disabled: true diff --git a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-activation-limit/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable-inheritance/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-continuable/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json index ef3f413d82..ed6f50a025 100644 --- a/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-diagnostic/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -393,7 +364,7 @@ }, { "name": "subagent_dsh_sdk", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages.", "parameters": { "type": "object", "properties": { @@ -408,18 +379,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -430,7 +389,7 @@ }, { "name": "subagent_dsh_sdk_second", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages.", "parameters": { "type": "object", "properties": { @@ -445,18 +404,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-list-agents/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json +++ b/snapshots/sdk/subagent-send-message/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/text-turn/tool-schemas.expected.json b/snapshots/sdk/text-turn/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/sdk/text-turn/tool-schemas.expected.json +++ b/snapshots/sdk/text-turn/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/sdk/tool-error-details/notifications.expected.jsonl b/snapshots/sdk/tool-error-details/notifications.expected.jsonl index 8d30d3d16c..e6d6e7fde2 100644 --- a/snapshots/sdk/tool-error-details/notifications.expected.jsonl +++ b/snapshots/sdk/tool-error-details/notifications.expected.jsonl @@ -6,7 +6,7 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"system/message","seq":7,"time":0,"data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":8,"time":0,"data":{"content":[{"type":"text","text":"Exercise a native denial and a caught PTC inner denial, then reply ERROR_DETAILS_OK."}],"source":{"kind":"user"},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"user/message","seq":9,"time":0,"data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{sessionId}}"},"surfaceOp":"append"}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":10,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/context","seq":11,"time":0,"data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"session/title","seq":12,"time":0,"data":{"title":"Exercise a native denial and","messageSeqs":[8],"source":{"kind":"fallback"}}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":13,"time":0,"data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"denied-native","name":"bash","args":["{\"command\":\"echo AUTO_DENIED\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} @@ -16,7 +16,7 @@ {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"step/start","seq":17,"time":0,"data":{"turn":1,"step":2}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"system/message","seq":18,"time":0,"data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{sessionId}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"request/header","seq":19,"time":0,"data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["run_code"]},"reason":"change","startsSeries":true}}}} -{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":20,"time":0,"data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"list_subagent_models"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":19},"surfaceOp":"append"}}} +{"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"developer/message","seq":20,"time":0,"data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{sessionId}}"},"headerSeq":19},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"assistant/message","seq":21,"time":0,"data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{sessionId}}"},"stream":[{"type":"chunk","time":0,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":0,"index":0,"dt":[],"id":"denied-code","name":"run_code","args":["{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"]},{"type":"chunk","time":0,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}}},{"type":"chunk","time":0,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/call","seq":22,"time":0,"data":{"turn":1,"step":2,"callId":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}}}} {"method":"session.event","params":{"sessionId":"{{sessionId}}","event":{"type":"tool/ptc-dispatch-start","seq":23,"time":0,"data":{"rootCallId":"denied-code","parentCallId":"denied-code","subCallId":"denied-code:ptc:1","name":"bash","arguments":{"command":"echo AUTO_DENIED"}}}}} diff --git a/snapshots/sdk/tool-error-details/session.v4.jsonl b/snapshots/sdk/tool-error-details/session.v4.jsonl index 806aa48ff4..871095019b 100644 --- a/snapshots/sdk/tool-error-details/session.v4.jsonl +++ b/snapshots/sdk/tool-error-details/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Exercise a native denial and a caught PTC inner denial, then reply ERROR_DETAILS_OK."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Exercise a native denial and","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"stream":[{"type":"chunk","time":1791315723125,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315723125,"index":0,"dt":[],"id":"denied-native","name":"bash","args":["{\"command\":\"echo AUTO_DENIED\"}"]},{"type":"chunk","time":1791315723125,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-native","name":"bash","arguments":"{\"command\":\"echo AUTO_DENIED\"}"}}},{"type":"chunk","time":1791315723125,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -19,7 +19,7 @@ {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["run_code"]},"reason":"change","startsSeries":true}} -{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"list_subagent_models"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} +{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1791315723151,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315723151,"index":0,"dt":[],"id":"denied-code","name":"run_code","args":["{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"]},{"type":"chunk","time":1791315723151,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}}},{"type":"chunk","time":1791315723151,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"denied-code","name":"run_code","arguments":"{\"code\":\"try { await tools.bash({command: \\\"echo AUTO_DENIED\\\"}) } catch (error) { return error.message }\",\"description\":\"Catch one denied inner call\"}"}} {"type":"tool/ptc-dispatch-start","data":{"rootCallId":"denied-code","parentCallId":"denied-code","subCallId":"denied-code:ptc:1","name":"bash","arguments":{"command":"echo AUTO_DENIED"}}} diff --git a/snapshots/sdk/tool-error-details/system-prompt.expected.md b/snapshots/sdk/tool-error-details/system-prompt.expected.md index 49a6727bc7..c7bcdec744 100644 --- a/snapshots/sdk/tool-error-details/system-prompt.expected.md +++ b/snapshots/sdk/tool-error-details/system-prompt.expected.md @@ -173,13 +173,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -206,7 +199,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -214,12 +207,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -453,7 +440,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/sdk/tool-error-details/tool-schemas.expected.json b/snapshots/sdk/tool-error-details/tool-schemas.expected.json index 628380a1d5..642da62fb7 100644 --- a/snapshots/sdk/tool-error-details/tool-schemas.expected.json +++ b/snapshots/sdk/tool-error-details/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/advanced-toolchain/session.v4.jsonl b/snapshots/session/advanced-toolchain/session.v4.jsonl index b9591d4802..c0f1a22b8d 100644 --- a/snapshots/session/advanced-toolchain/session.v4.jsonl +++ b/snapshots/session/advanced-toolchain/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Run this advanced flow exactly once: list runtime inspect providers; discover providers through run_code when available; delegate once to a direct spawn child; run one workflow that delegates to another spawn child; list providers again; then reply with exactly ADVANCED_HEADLESS_OK."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Run this advanced flow exactly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-inspect-before","name":"cordis_inspect_list","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1791315715021,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315715022,"index":0,"dt":[],"id":"advanced-inspect-before","name":"cordis_inspect_list","args":["{}"]},{"type":"chunk","time":1791315715022,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-inspect-before","name":"cordis_inspect_list","arguments":"{}"}}},{"type":"chunk","time":1791315715022,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1791315715022,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -19,7 +19,7 @@ {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["run_code"]},"reason":"change","startsSeries":true}} -{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"cordis_inspect_list"},{"type":"tool-removal","toolName":"cordis_inspect_query"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"list_subagent_models"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} +{"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"run_code"},{"type":"tool-removal","toolName":"bash"},{"type":"tool-removal","toolName":"cordis_inspect_list"},{"type":"tool-removal","toolName":"cordis_inspect_query"},{"type":"tool-removal","toolName":"create_goal"},{"type":"tool-removal","toolName":"edit"},{"type":"tool-removal","toolName":"exit_plan_mode"},{"type":"tool-removal","toolName":"get_goal"},{"type":"tool-removal","toolName":"glob"},{"type":"tool-removal","toolName":"grep"},{"type":"tool-removal","toolName":"interrupt_agent"},{"type":"tool-removal","toolName":"job_kill"},{"type":"tool-removal","toolName":"job_list"},{"type":"tool-removal","toolName":"job_output"},{"type":"tool-removal","toolName":"list_agents"},{"type":"tool-removal","toolName":"read"},{"type":"tool-removal","toolName":"read_image"},{"type":"tool-removal","toolName":"send_message"},{"type":"tool-removal","toolName":"skill"},{"type":"tool-removal","toolName":"subagent"},{"type":"tool-removal","toolName":"subagent_fork"},{"type":"tool-removal","toolName":"todo_write"},{"type":"tool-removal","toolName":"update_goal"},{"type":"tool-removal","toolName":"web_fetch"},{"type":"tool-removal","toolName":"web_search"},{"type":"tool-removal","toolName":"workflow"},{"type":"tool-removal","toolName":"working_directory"},{"type":"tool-removal","toolName":"write"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-code","name":"run_code","arguments":"{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1791315715055,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315715055,"index":0,"dt":[],"id":"advanced-code","name":"run_code","args":["{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"]},{"type":"chunk","time":1791315715055,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-code","name":"run_code","arguments":"{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"}}},{"type":"chunk","time":1791315715055,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1791315715055,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"advanced-code","name":"run_code","arguments":"{\"code\":\"const inspected = await tools.cordis_inspect_list({}); return inspected.providers.map(provider => provider.id);\",\"description\":\"Discover runtime APIs through PTC\"}"}} {"type":"tool/ptc-dispatch-start","data":{"rootCallId":"advanced-code","parentCallId":"advanced-code","subCallId":"advanced-code:ptc:1","name":"cordis_inspect_list","arguments":{}}} @@ -28,8 +28,8 @@ {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} {"type":"system/message","data":{"turn":1,"step":3,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:10}}"}},"sourceEventSeqs":[18],"surfaceOp":{"op":"replace","startSeq":18,"endSeq":18}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} -{"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"cordis_inspect_list"},{"type":"tool-addition","toolName":"cordis_inspect_query"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"list_subagent_models"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"},{"type":"tool-removal","toolName":"run_code"}],"role":"developer","id":"{{message:11}}"},"headerSeq":29},"surfaceOp":"append"} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":["bash","cordis_inspect_list","cordis_inspect_query","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} +{"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"bash"},{"type":"tool-addition","toolName":"cordis_inspect_list"},{"type":"tool-addition","toolName":"cordis_inspect_query"},{"type":"tool-addition","toolName":"create_goal"},{"type":"tool-addition","toolName":"edit"},{"type":"tool-addition","toolName":"exit_plan_mode"},{"type":"tool-addition","toolName":"get_goal"},{"type":"tool-addition","toolName":"glob"},{"type":"tool-addition","toolName":"grep"},{"type":"tool-addition","toolName":"interrupt_agent"},{"type":"tool-addition","toolName":"job_kill"},{"type":"tool-addition","toolName":"job_list"},{"type":"tool-addition","toolName":"job_output"},{"type":"tool-addition","toolName":"list_agents"},{"type":"tool-addition","toolName":"read"},{"type":"tool-addition","toolName":"read_image"},{"type":"tool-addition","toolName":"send_message"},{"type":"tool-addition","toolName":"skill"},{"type":"tool-addition","toolName":"subagent"},{"type":"tool-addition","toolName":"subagent_fork"},{"type":"tool-addition","toolName":"todo_write"},{"type":"tool-addition","toolName":"update_goal"},{"type":"tool-addition","toolName":"web_fetch"},{"type":"tool-addition","toolName":"web_search"},{"type":"tool-addition","toolName":"workflow"},{"type":"tool-addition","toolName":"working_directory"},{"type":"tool-addition","toolName":"write"},{"type":"tool-removal","toolName":"run_code"}],"role":"developer","id":"{{message:11}}"},"headerSeq":29},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"tool-call","id":"advanced-direct-child","name":"subagent","arguments":"{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:12}}"},"usage":{"inputTokens":3,"outputTokens":3},"stream":[{"type":"chunk","time":1791315715200,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791315715200,"index":0,"dt":[],"id":"advanced-direct-child","name":"subagent","args":["{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"]},{"type":"chunk","time":1791315715200,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"advanced-direct-child","name":"subagent","arguments":"{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"}}},{"type":"chunk","time":1791315715200,"chunk":{"type":"usage","usage":{"inputTokens":3,"outputTokens":3}}},{"type":"chunk","time":1791315715200,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":3,"callId":"advanced-direct-child","name":"subagent","arguments":"{\"description\":\"Check direct child\",\"prompt\":\"Reply with exactly DIRECT_CHILD_OK and nothing else.\",\"run_in_background\":false}"}} {"type":"subagent/catalog","data":{"version":0,"childId":"{{session:2}}","childCreatedAt":1791315715221,"mode":"continuable","label":"Check direct child"}} diff --git a/snapshots/session/advanced-toolchain/system-prompt.expected.md b/snapshots/session/advanced-toolchain/system-prompt.expected.md index 2cd6250c0f..c97a5b8fcb 100644 --- a/snapshots/session/advanced-toolchain/system-prompt.expected.md +++ b/snapshots/session/advanced-toolchain/system-prompt.expected.md @@ -186,13 +186,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -219,7 +212,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -227,12 +220,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -468,7 +455,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/advanced-toolchain/tool-schemas.expected.json b/snapshots/session/advanced-toolchain/tool-schemas.expected.json index 89dd15e503..14756944ff 100644 --- a/snapshots/session/advanced-toolchain/tool-schemas.expected.json +++ b/snapshots/session/advanced-toolchain/tool-schemas.expected.json @@ -301,23 +301,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -397,7 +380,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -412,18 +395,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -1043,23 +1014,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1139,7 +1093,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1154,18 +1108,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/agent-instructions/tool-schemas.expected.json b/snapshots/session/agent-instructions/tool-schemas.expected.json index ffcb7cdbcb..00e71002a7 100644 --- a/snapshots/session/agent-instructions/tool-schemas.expected.json +++ b/snapshots/session/agent-instructions/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -922,23 +893,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1018,7 +972,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1033,18 +987,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/background-confinement-failure/tool-schemas.expected.json b/snapshots/session/background-confinement-failure/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/background-confinement-failure/tool-schemas.expected.json +++ b/snapshots/session/background-confinement-failure/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/both-mode-turn/system-prompt.expected.md b/snapshots/session/both-mode-turn/system-prompt.expected.md index da16df0b27..d58e71a227 100644 --- a/snapshots/session/both-mode-turn/system-prompt.expected.md +++ b/snapshots/session/both-mode-turn/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -420,7 +407,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json index f96956dd76..cd473bea70 100644 --- a/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-chrome-devtools-mcp/tool-schemas.expected.json @@ -300,23 +300,6 @@ ] } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "mcp__chrome-devtools-mcp__take_screenshot", "description": "Take a screenshot of the page or element.", @@ -463,7 +446,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -478,18 +461,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json index 2fc2120472..2d24d7c17d 100644 --- a/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json +++ b/snapshots/session/browser-use-playwright-mcp/tool-schemas.expected.json @@ -300,23 +300,6 @@ ] } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "mcp__playwright-mcp__browser_take_screenshot", "description": "Take a screenshot of the current page. You can't perform actions based on the screenshot, use browser_snapshot for actions.", @@ -465,7 +448,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -480,18 +463,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json index d8f286c03c..e1acbc8206 100644 --- a/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json +++ b/snapshots/session/browser-use-stagehand-native/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -569,7 +552,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -584,18 +567,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/claude-code-mods/tool-schemas.expected.json b/snapshots/session/claude-code-mods/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/claude-code-mods/tool-schemas.expected.json +++ b/snapshots/session/claude-code-mods/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/compaction-output-reserve/cordis.snapshot.yml b/snapshots/session/compaction-output-reserve/cordis.snapshot.yml index b9294ef273..3dce48221f 100644 --- a/snapshots/session/compaction-output-reserve/cordis.snapshot.yml +++ b/snapshots/session/compaction-output-reserve/cordis.snapshot.yml @@ -1,5 +1,5 @@ # With 1,800 tokens of headroom, a 1,000-token output reservation lowers the -# trigger from 7,470 to 6,470, between the history and its condensed replacement. +# trigger from 6,782 to 5,782, between the history and its condensed replacement. - id: agent-instructions disabled: true @@ -31,7 +31,7 @@ name: DeepSeek models: - id: deepseek-v4-flash - contextWindow: 9009 + contextWindow: 8582 defaultMaxTokens: 1000 - id: compaction-working-directory name: '../../../packages/test-support/session-snapshot/tests/fixtures/compaction-working-directory.ts' diff --git a/snapshots/session/compaction-output-reserve/cordis.yml b/snapshots/session/compaction-output-reserve/cordis.yml index e970dde73c..a8d2088083 100644 --- a/snapshots/session/compaction-output-reserve/cordis.yml +++ b/snapshots/session/compaction-output-reserve/cordis.yml @@ -7,7 +7,7 @@ config: models: - id: deepseek-v4-flash - contextWindow: 9009 + contextWindow: 8582 maxTokens: 1000 - id: compaction-basic diff --git a/snapshots/session/compaction-output-reserve/session.v4.jsonl b/snapshots/session/compaction-output-reserve/session.v4.jsonl index 51414cb663..6b7fb48868 100644 --- a/snapshots/session/compaction-output-reserve/session.v4.jsonl +++ b/snapshots/session/compaction-output-reserve/session.v4.jsonl @@ -11,7 +11,7 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Establish a durable compaction premise before continuing. Record every part of this historical evidence: the snapshot uses keyless replay; persistence uses JSONL; the assembled headless application loads its real Cordis composition; model-visible inputs remain logged; tool calls and results remain paired and ordered; output reservation reduces the message budget before proactive compaction; compaction opens with compaction/start and closes with compaction/end; a successful auxiliary summary records compaction/summary source metadata; the replacement surface shadows only an older balanced range; the checkpoint remains smaller than the history it replaces; the newest tool result remains verbatim; the next request sees that checkpoint; the final response proves the same turn continued; deterministic snapshot evidence stays separate from the live-provider smoke; and no external API key is needed. Emit one alpha marker through bash, then finish the task after proactive compaction with the exact words PROACTIVE COMPACTION COMPLETE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"/tmp\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"/tmp\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":1000},"adapterDefaults":{"maxTokens":true},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9009}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":8582}} {"type":"session/title","data":{"title":"Establish a durable compaction premise","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":24,"outputTokens":6},"stream":[{"type":"chunk","time":1791427805327,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791427805327,"index":0,"dt":[],"id":"call_compaction_marker","name":"bash","args":["{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"]},{"type":"chunk","time":1791427805327,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}}},{"type":"chunk","time":1791427805327,"chunk":{"type":"usage","usage":{"inputTokens":24,"outputTokens":6}}},{"type":"chunk","time":1791427805327,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}} diff --git a/snapshots/session/compaction-recovery/tool-schemas.expected.json b/snapshots/session/compaction-recovery/tool-schemas.expected.json index ffcb7cdbcb..00e71002a7 100644 --- a/snapshots/session/compaction-recovery/tool-schemas.expected.json +++ b/snapshots/session/compaction-recovery/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -922,23 +893,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1018,7 +972,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1033,18 +987,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml b/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml index 342f2c3574..765a3f7e8a 100644 --- a/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml +++ b/snapshots/session/compaction-summary-headroom/cordis.snapshot.yml @@ -1,5 +1,5 @@ # The summary output cap inherits the 1,800-token headroom. A 1,000-token -# request output reservation lowers the proactive trigger to 6,470 tokens. +# request output reservation lowers the proactive trigger to 5,782 tokens. - id: agent-instructions disabled: true @@ -30,7 +30,7 @@ name: DeepSeek models: - id: deepseek-v4-flash - contextWindow: 9009 + contextWindow: 8582 defaultMaxTokens: 1000 - id: compaction-working-directory name: '../../../packages/test-support/session-snapshot/tests/fixtures/compaction-working-directory.ts' diff --git a/snapshots/session/compaction-summary-headroom/cordis.yml b/snapshots/session/compaction-summary-headroom/cordis.yml index 1897460f05..f65c90899e 100644 --- a/snapshots/session/compaction-summary-headroom/cordis.yml +++ b/snapshots/session/compaction-summary-headroom/cordis.yml @@ -7,7 +7,7 @@ config: models: - id: deepseek-v4-flash - contextWindow: 9009 + contextWindow: 8582 maxTokens: 1000 - id: compaction-basic diff --git a/snapshots/session/compaction-summary-headroom/session.v4.jsonl b/snapshots/session/compaction-summary-headroom/session.v4.jsonl index aafe00a914..b3ba4ad911 100644 --- a/snapshots/session/compaction-summary-headroom/session.v4.jsonl +++ b/snapshots/session/compaction-summary-headroom/session.v4.jsonl @@ -11,7 +11,7 @@ {"type":"user/message","data":{"content":[{"type":"text","text":"Establish a durable compaction premise before continuing. Record every part of this historical evidence: the snapshot uses keyless replay; persistence uses JSONL; the assembled headless application loads its real Cordis composition; model-visible inputs remain logged; tool calls and results remain paired and ordered; output reservation reduces the message budget before proactive compaction; compaction opens with compaction/start and closes with compaction/end; a successful auxiliary summary records compaction/summary source metadata; the replacement surface shadows only an older balanced range; the checkpoint remains smaller than the history it replaces; the newest tool result remains verbatim; the next request sees that checkpoint; the final response proves the same turn continued; deterministic snapshot evidence stays separate from the live-provider smoke; and no external API key is needed. Emit one alpha marker through bash, then finish the task after proactive compaction with the exact words PROACTIVE COMPACTION COMPLETE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"/tmp\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"/tmp\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":1000},"adapterDefaults":{"maxTokens":true},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":9009}} +{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":8582}} {"type":"session/title","data":{"title":"Establish a durable compaction premise","messageSeqs":[9],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":24,"outputTokens":6},"stream":[{"type":"chunk","time":1791427807038,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791427807038,"index":0,"dt":[],"id":"call_compaction_marker","name":"bash","args":["{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"]},{"type":"chunk","time":1791427807038,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}}},{"type":"chunk","time":1791427807038,"chunk":{"type":"usage","usage":{"inputTokens":24,"outputTokens":6}}},{"type":"chunk","time":1791427807038,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_compaction_marker","name":"bash","arguments":"{\"command\":\"printf 'alpha\\n'\",\"description\":\"Emit compaction premise marker\"}"}} diff --git a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json index 4a4cca52f0..4b839adb31 100644 --- a/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-mcp/tool-schemas.expected.json @@ -300,23 +300,6 @@ ] } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "mcp__cua-driver-mcp__disconnect", "description": "Disconnect the fixture.", @@ -442,7 +425,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -457,18 +440,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json index 3957dbc2f8..802ed2fe0d 100644 --- a/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json +++ b/snapshots/session/computer-use-cua-driver-native/tool-schemas.expected.json @@ -306,23 +306,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -402,7 +385,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -417,18 +400,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json index 2e5f4fc018..5d880fdde7 100644 --- a/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json +++ b/snapshots/session/cordis-inspect-jsdoc/tool-schemas.expected.json @@ -301,23 +301,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -397,7 +380,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -412,18 +395,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-messages-system-prompt/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json +++ b/snapshots/session/deepseek-protocol-system-prompt/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl b/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl index 53e2ccd74b..cff0e5a329 100644 --- a/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl +++ b/snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","systemPromptUpdate":"in-history"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790149644504,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644504,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790149644504,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -18,14 +18,14 @@ {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1790149644534,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644534,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790149644534,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790149644535,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:9}}"}},"sourceEventSeqs":[22],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:10}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:11}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790149644557,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790149644557,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790149644557,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/session/dynamic-tool-updates/session.v4.jsonl b/snapshots/session/dynamic-tool-updates/session.v4.jsonl index 8472750d16..aee79405da 100644 --- a/snapshots/session/dynamic-tool-updates/session.v4.jsonl +++ b/snapshots/session/dynamic-tool-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790049391235,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391235,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790049391235,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790049391235,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -17,14 +17,14 @@ {"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_task_read"},"toolCallId":"call_task_read","content":[{"type":"text","text":"{{cwd}}/task.txt\nfile\n\n1: Reply with the single word DONE after reading this file.\n\n(End of file - total 1 lines)\n"}],"isError":false,"id":"{{message:5}}"},"meta":{"path":"{{cwd}}/task.txt","offset":1,"lines":[{"number":1,"text":"Reply with the single word DONE after reading this file."}],"totalLines":1}},"sourceEventSeqs":[14],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:6}}"},"headerSeq":18},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:7}}"},"stream":[{"type":"chunk","time":1790049391255,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790049391255,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790049391255,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790049391255,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:8}}"}},"sourceEventSeqs":[21],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"high"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:9}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:10}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790049391270,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790049391270,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790049391270,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790049391270,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json index 798fc8024b..87bccafb3f 100644 --- a/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json +++ b/snapshots/session/dynamic-tool-updates/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -922,23 +893,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1026,7 +980,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1041,18 +995,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -1591,23 +1533,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -1687,7 +1612,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1702,18 +1627,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json +++ b/snapshots/session/foreground-confinement-timeout/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json index 03764bcd7e..3fb3980c3b 100644 --- a/snapshots/session/fs-glob-sampling/tool-schemas.expected.json +++ b/snapshots/session/fs-glob-sampling/tool-schemas.expected.json @@ -180,23 +180,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -260,7 +243,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -275,18 +258,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/lsp-definition/tool-schemas.expected.json b/snapshots/session/lsp-definition/tool-schemas.expected.json index 2bc62b0f70..76087bb96f 100644 --- a/snapshots/session/lsp-definition/tool-schemas.expected.json +++ b/snapshots/session/lsp-definition/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "lsp", "description": "Query a language server for precise code navigation. operation is one of goToDefinition, findReferences, goToImplementation, hover. line and character are one-based UTF-16 cursor coordinates. findReferences includes the declaration.", @@ -393,7 +376,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -408,18 +391,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md index b81f602015..e9b26ca33b 100644 --- a/snapshots/session/mcp-resources-ptc/system-prompt.expected.md +++ b/snapshots/session/mcp-resources-ptc/system-prompt.expected.md @@ -162,13 +162,6 @@ interface ToolArgsMap { /** Continuation cursor returned by this server. */ cursor?: string; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -202,7 +195,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -210,12 +203,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -451,7 +438,6 @@ interface ToolOutputMap { })[]; list_mcp_resource_templates: JsonValue; list_mcp_resources: JsonValue; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/mcp-resources/tool-schemas.expected.json b/snapshots/session/mcp-resources/tool-schemas.expected.json index fa72eb6521..4e5e9d557c 100644 --- a/snapshots/session/mcp-resources/tool-schemas.expected.json +++ b/snapshots/session/mcp-resources/tool-schemas.expected.json @@ -300,23 +300,6 @@ ] } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -417,7 +400,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -432,18 +415,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json index fe1f0ce7a4..637c069b74 100644 --- a/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json +++ b/snapshots/session/multimodal-spill-ends/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "mcp__fixture__inspect", "description": "Read a fixed window snapshot.", @@ -375,7 +358,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -390,18 +373,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/office-progressive-native/tool-schemas.expected.json b/snapshots/session/office-progressive-native/tool-schemas.expected.json index 36be4284f4..79d6891abc 100644 --- a/snapshots/session/office-progressive-native/tool-schemas.expected.json +++ b/snapshots/session/office-progressive-native/tool-schemas.expected.json @@ -215,23 +215,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -311,7 +294,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -326,18 +309,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/office-progressive-ptc/system-prompt.expected.md b/snapshots/session/office-progressive-ptc/system-prompt.expected.md index e9119ddc55..9d427282d8 100644 --- a/snapshots/session/office-progressive-ptc/system-prompt.expected.md +++ b/snapshots/session/office-progressive-ptc/system-prompt.expected.md @@ -119,13 +119,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -152,7 +145,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -160,12 +153,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -363,7 +350,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/office-skills/tool-schemas.expected.json b/snapshots/session/office-skills/tool-schemas.expected.json index 36be4284f4..79d6891abc 100644 --- a/snapshots/session/office-skills/tool-schemas.expected.json +++ b/snapshots/session/office-skills/tool-schemas.expected.json @@ -215,23 +215,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -311,7 +294,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -326,18 +309,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json index 4195c5187a..d1b4eef213 100644 --- a/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/persistent-pwsh-tool-turn/tool-schemas.expected.json @@ -135,23 +135,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "pwsh", "description": "Run commands in a persistent PowerShell shell. State, including the current directory and exported environment variables, persists across calls for this agent.", @@ -231,7 +214,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -246,18 +229,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl b/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl index 6e26da9e41..843a75e88d 100644 --- a/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl +++ b/snapshots/session/pi-ai-conversation-updates/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Read task.txt, call the newly available snapshot_ping tool once, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"moonshotai","model":"kimi-k3","contextWindow":1048576,"systemPromptUpdate":"in-history"}} {"type":"session/title","data":{"title":"Read task.txt, call the newly","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}],"source":{"kind":"model","provider":"moonshotai","model":"kimi-k3"},"id":"{{message:4}}"},"usage":{"inputTokens":10,"outputTokens":5},"stream":[{"type":"chunk","time":1790149644504,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644504,"index":0,"dt":[],"id":"call_task_read","name":"read","args":["{\"file_path\":\"task.txt\"}"]},{"type":"chunk","time":1790149644504,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_task_read","name":"read","arguments":"{\"file_path\":\"task.txt\"}"}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"usage","usage":{"inputTokens":10,"outputTokens":5}}},{"type":"chunk","time":1790149644504,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} @@ -18,14 +18,14 @@ {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","snapshot_ping","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}],"source":{"kind":"model","provider":"moonshotai","model":"kimi-k3"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1790149644534,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790149644534,"index":0,"dt":[],"id":"call_dynamic_ping","name":"snapshot_ping","args":["{}"]},{"type":"chunk","time":1790149644534,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}}},{"type":"chunk","time":1790149644535,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_dynamic_ping","name":"snapshot_ping","arguments":"{}"}} {"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_dynamic_ping"},"toolCallId":"call_dynamic_ping","content":[{"type":"text","text":"pong"}],"isError":false,"id":"{{message:9}}"}},"sourceEventSeqs":[22],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} -{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} +{"type":"request/header","data":{"header":{"config":{"provider":"moonshotai","model":"kimi-k3"},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change"}} {"type":"developer/message","data":{"turn":1,"step":3,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-removal","toolName":"snapshot_ping"}],"role":"developer","id":"{{message:10}}"}},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"moonshotai","model":"kimi-k3"},"id":"{{message:11}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1790149644557,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1790149644557,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1790149644557,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1790149644557,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} diff --git a/snapshots/session/plugin-manager-mcp/session.v4.jsonl b/snapshots/session/plugin-manager-mcp/session.v4.jsonl index 0036cc1272..794d19b23f 100644 --- a/snapshots/session/plugin-manager-mcp/session.v4.jsonl +++ b/snapshots/session/plugin-manager-mcp/session.v4.jsonl @@ -9,7 +9,7 @@ {"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Use plugin_manager with action set_plugin, target include:demo, enabled true to activate the configured MCP entry. Then call its newly discovered mcp__demo__ping tool and report the result. Do not use other tools."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} {"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_subagent_models","plugin_manager","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","plugin_manager","read","read_image","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"initial"}} {"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash","contextWindow":1000000}} {"type":"session/title","data":{"title":"Use plugin_manager with action set_plugi","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"reasoning","text":"The user wants me to activate a plugin via plugin_manager, then call an MCP tool. Let me do the plugin_manager call first.\n\nNote: they say \"Do not use other tools\" other than plugin_manager and the mcp tool. But I need to first enable it. Let's call plugin_manager set_plugin.\n\nLet me also list plugins first? The instruction says use plugin_manager with action set_plugin, target include:demo, enabled true. Let's just do that."},{"type":"tool-call","id":"call_00_xLhr9wia3A5m7wuSRWwD0436","name":"plugin_manager","arguments":"{\"action\": \"set_plugin\", \"target\": \"include:demo\", \"enabled\": true}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"75d99e21-1d47-4a33-bb88-340a1e84b749"},{"type":"tool-call"}]}},"id":"{{message:4}}"},"usage":{"inputTokens":7220,"outputTokens":174,"cacheReadTokens":0,"cacheWriteTokens":0,"totalTokens":7394},"stream":[{"type":"chunk","time":1789984418848,"chunk":{"type":"block-start","index":0,"blockType":"reasoning"}},{"type":"reasoning-chunks","time0":1789984418848,"index":0,"dt":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"texts":["The"," user"," wants"," me"," to"," activate"," a"," plugin"," via"," plugin","_manager",","," then"," call"," an"," M","CP"," tool","."," Let"," me"," do"," the"," plugin","_manager"," call"," first",".\n\n","Note",":"," they"," say"," \"","Do"," not"," use"," other"," tools","\""," other"," than"," plugin","_manager"," and"," the"," m","cp"," tool","."," But"," I"," need"," to"," first"," enable"," it","."," Let","'s"," call"," plugin","_manager"," set","_","plugin",".\n\n","Let"," me"," also"," list"," plugins"," first","?"," The"," instruction"," says"," use"," plugin","_manager"," with"," action"," set","_","plugin",","," target"," include",":","demo",","," enabled"," true","."," Let","'s"," just"," do"," that","."]},{"type":"chunk","time":1789984418849,"chunk":{"type":"block-end","index":0,"block":{"type":"reasoning","text":"The user wants me to activate a plugin via plugin_manager, then call an MCP tool. Let me do the plugin_manager call first.\n\nNote: they say \"Do not use other tools\" other than plugin_manager and the mcp tool. But I need to first enable it. Let's call plugin_manager set_plugin.\n\nLet me also list plugins first? The instruction says use plugin_manager with action set_plugin, target include:demo, enabled true. Let's just do that."}}},{"type":"chunk","time":1789984418849,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1789984418849,"index":1,"dt":[],"id":"call_00_xLhr9wia3A5m7wuSRWwD0436","name":"plugin_manager","args":[""]},{"type":"tool-call-chunks","time0":1789984418849,"index":1,"dt":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"id":"call_00_xLhr9wia3A5m7wuSRWwD0436","args":["{","\"","action","\"",": ","\"","set","_","plugin","\"",", ","\"","target","\"",": ","\"","include",":","demo","\"",", ","\"","en","abled","\"",": ","true","}"]},{"type":"chunk","time":1789984418849,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"call_00_xLhr9wia3A5m7wuSRWwD0436","name":"plugin_manager","arguments":"{\"action\": \"set_plugin\", \"target\": \"include:demo\", \"enabled\": true}"}}},{"type":"chunk","time":1789984418849,"chunk":{"type":"usage","usage":{"inputTokens":7220,"outputTokens":174,"cacheReadTokens":0,"cacheWriteTokens":0,"totalTokens":7394}}},{"type":"chunk","time":1789984418849,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"75d99e21-1d47-4a33-bb88-340a1e84b749"},{"type":"tool-call"}]}}}]},"surfaceOp":"append"} @@ -18,7 +18,7 @@ {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"system/message","data":{"turn":1,"step":2,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:6}}"}},"sourceEventSeqs":[7],"surfaceOp":{"op":"replace","startSeq":7,"endSeq":7}} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_mcp_resource_templates","list_mcp_resources","list_subagent_models","mcp__demo__ping","plugin_manager","read","read_image","read_mcp_resource","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} +{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash","maxTokens":256000,"reasoningEffort":"max"},"adapterDefaults":{"reasoningEffort":true,"maxTokens":true},"tools":["bash","create_goal","edit","exit_plan_mode","get_goal","glob","grep","interrupt_agent","job_kill","job_list","job_output","list_agents","list_mcp_resource_templates","list_mcp_resources","mcp__demo__ping","plugin_manager","read","read_image","read_mcp_resource","send_message","skill","subagent","subagent_fork","todo_write","update_goal","web_fetch","web_search","workflow","working_directory","write"]},"reason":"change","startsSeries":true}} {"type":"developer/message","data":{"turn":1,"step":2,"message":{"source":{"kind":"tool-registry"},"content":[{"type":"tool-addition","toolName":"list_mcp_resource_templates"},{"type":"tool-addition","toolName":"list_mcp_resources"},{"type":"tool-addition","toolName":"mcp__demo__ping"},{"type":"tool-addition","toolName":"read_mcp_resource"}],"role":"developer","id":"{{message:7}}"},"headerSeq":19},"surfaceOp":"append"} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"reasoning","text":"The plugin was enabled. Now call mcp__demo__ping.\n\nWait — the tool was already listed in my tools before enabling. Let's call it anyway."},{"type":"tool-call","id":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","arguments":"{}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash","replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"b1ec1efe-4be6-4836-a10c-e5d6bfdce520"},{"type":"tool-call"}]}},"id":"{{message:8}}"},"usage":{"inputTokens":7035,"outputTokens":59,"cacheReadTokens":768,"cacheWriteTokens":0,"totalTokens":7862},"stream":[{"type":"chunk","time":1789984418998,"chunk":{"type":"block-start","index":0,"blockType":"reasoning"}},{"type":"reasoning-chunks","time0":1789984418998,"index":0,"dt":[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0],"texts":["The"," plugin"," was"," enabled","."," Now"," call"," m","cp","__","demo","__","ping",".\n\n","Wait"," —"," the"," tool"," was"," already"," listed"," in"," my"," tools"," before"," enabling","."," Let","'s"," call"," it"," anyway","."]},{"type":"chunk","time":1789984418998,"chunk":{"type":"block-end","index":0,"block":{"type":"reasoning","text":"The plugin was enabled. Now call mcp__demo__ping.\n\nWait — the tool was already listed in my tools before enabling. Let's call it anyway."}}},{"type":"chunk","time":1789984418998,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1789984418998,"index":1,"dt":[],"id":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","args":[""]},{"type":"tool-call-chunks","time0":1789984418998,"index":1,"dt":[],"id":"call_00_miuKfLCTz5Cb38gdp8kK5186","args":["{}"]},{"type":"chunk","time":1789984418998,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","arguments":"{}"}}},{"type":"chunk","time":1789984418998,"chunk":{"type":"usage","usage":{"inputTokens":7035,"outputTokens":59,"cacheReadTokens":768,"cacheWriteTokens":0,"totalTokens":7862}}},{"type":"chunk","time":1789984418998,"chunk":{"type":"finish","reason":{"kind":"tool-calls"},"replayState":{"response":{"kind":"deepseek-messages","version":1,"model":"deepseek-v4-flash"},"blocks":[{"type":"reasoning","signature":"b1ec1efe-4be6-4836-a10c-e5d6bfdce520"},{"type":"tool-call"}]}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_00_miuKfLCTz5Cb38gdp8kK5186","name":"mcp__demo__ping","arguments":"{}"}} diff --git a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json index ca83d406e0..f44f52640a 100644 --- a/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager-mcp/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "plugin_manager", "description": "List plugins or bundles in the current profile, enable or disable them, install a bundle, or remove an installed bundle. Every action requires danger-full-access permission or approval for this call. Approval does not change the session permission mode. Changes affect every session in this profile. List first to obtain exact identifiers. Package installation can execute allowed build scripts. Live profiles apply changes immediately; startup profiles require restart. Incompatible DSH peer dependencies block installation and activation. Version exemptions risk crashes and data loss: warn the user and obtain explicit permission for the exact plugin and runtime versions before granting one.", @@ -417,7 +400,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -432,18 +415,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ @@ -1023,23 +994,6 @@ ] } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "mcp__demo__ping", "description": "Replies pong.", @@ -1210,7 +1164,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -1225,18 +1179,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/plugin-manager/tool-schemas.expected.json b/snapshots/session/plugin-manager/tool-schemas.expected.json index 62bdec5dc1..4596dae7a1 100644 --- a/snapshots/session/plugin-manager/tool-schemas.expected.json +++ b/snapshots/session/plugin-manager/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "plugin_manager", "description": "List plugins or bundles in the current profile, enable or disable them, install a bundle, or remove an installed bundle. Every action requires danger-full-access permission or approval for this call. Approval does not change the session permission mode. Changes affect every session in this profile. List first to obtain exact identifiers. Package installation can execute allowed build scripts. Live profiles apply changes immediately; startup profiles require restart. Incompatible DSH peer dependencies block installation and activation. Version exemptions risk crashes and data loss: warn the user and obtain explicit permission for the exact plugin and runtime versions before granting one.", @@ -417,7 +400,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -432,18 +415,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/product-subagent-both/tool-schemas.expected.json b/snapshots/session/product-subagent-both/tool-schemas.expected.json index 08ca2b1a53..2b36308f92 100644 --- a/snapshots/session/product-subagent-both/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-both/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/product-subagent-codex/tool-schemas.expected.json b/snapshots/session/product-subagent-codex/tool-schemas.expected.json index 49f6480eb1..d9f3cfa9d0 100644 --- a/snapshots/session/product-subagent-codex/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-codex/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json index 34004323e2..11f07197a2 100644 --- a/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/product-subagent-result-diagnostic/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/provider-cwd/tool-schemas.expected.json b/snapshots/session/provider-cwd/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/provider-cwd/tool-schemas.expected.json +++ b/snapshots/session/provider-cwd/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ptc-node-read-only/system-prompt.expected.md b/snapshots/session/ptc-node-read-only/system-prompt.expected.md index 1e964bdb4f..2801645046 100644 --- a/snapshots/session/ptc-node-read-only/system-prompt.expected.md +++ b/snapshots/session/ptc-node-read-only/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -424,7 +411,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-node-workspace/system-prompt.expected.md b/snapshots/session/ptc-node-workspace/system-prompt.expected.md index 1e964bdb4f..2801645046 100644 --- a/snapshots/session/ptc-node-workspace/system-prompt.expected.md +++ b/snapshots/session/ptc-node-workspace/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -424,7 +411,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-persistent/system-prompt.expected.md b/snapshots/session/ptc-persistent/system-prompt.expected.md index 1e964bdb4f..2801645046 100644 --- a/snapshots/session/ptc-persistent/system-prompt.expected.md +++ b/snapshots/session/ptc-persistent/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -424,7 +411,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-policy-change/system-prompt.expected.md b/snapshots/session/ptc-policy-change/system-prompt.expected.md index 65c682d835..19204006b3 100644 --- a/snapshots/session/ptc-policy-change/system-prompt.expected.md +++ b/snapshots/session/ptc-policy-change/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -422,7 +409,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-python-turn/system-prompt.expected.md b/snapshots/session/ptc-python-turn/system-prompt.expected.md index a6dfe8d41b..6e5a4c7a78 100644 --- a/snapshots/session/ptc-python-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-python-turn/system-prompt.expected.md @@ -287,13 +287,6 @@ class ListAgentsOutput2(TypedDict): parent: NotRequired[str] depth: NotRequired[float] -class ListSubagentModelsArgs(TypedDict): - # Registered LLM provider id. - provider: NotRequired[str] - # Exact model id; requires provider. - model: NotRequired[str] - # Additional keys beyond those declared are allowed. - class ReadArgs(TypedDict): # Path to read, resolved by the filesystem backend. file_path: str @@ -377,12 +370,6 @@ class SubagentArgs(TypedDict): description: str # The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. prompt: str - # Child LLM provider id. - provider: NotRequired[str] - # Model id within that provider. - model: NotRequired[str] - # Model-specific reasoning effort id. - reasoning_effort: NotRequired[str] # Additional keys beyond those declared are allowed. class SubagentOutput(TypedDict): @@ -548,8 +535,6 @@ class Tools(Protocol): """Read a background job: output since the previous read for stream jobs, or the result of a finished final-output job.""" async def list_agents(self, args: ListAgentsArgs) -> list[ListAgentsOutput1 | ListAgentsOutput2]: """List subagents you started, with their ids, labels, and status. running means it is working; inactive means it is not currently working. You will be notified when a subagent finishes; there is no need to keep checking its status. Use send_message to continue the conversation.""" - async def list_subagent_models(self, args: ListSubagentModelsArgs) -> str: - """Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.""" async def read(self, args: ReadArgs) -> ReadOutput: """Read a UTF-8 text file and return line-numbered content.""" async def read_image(self, args: ReadImageArgs) -> ReadImageOutput: @@ -559,7 +544,7 @@ class Tools(Protocol): async def skill(self, args: SkillArgs) -> SkillOutput: """Load the full instructions for a skill. Call it before acting on a task that names or clearly matches a skill in the session skill catalog.""" async def subagent(self, args: SubagentArgs) -> SubagentOutput: - """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.""" + """Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.""" async def subagent_fork(self, args: SubagentForkArgs) -> SubagentForkOutput: """Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.""" async def todo_write(self, args: TodoWriteArgs) -> TodoWriteOutput: diff --git a/snapshots/session/ptc-read-image/system-prompt.expected.md b/snapshots/session/ptc-read-image/system-prompt.expected.md index 4d6f11b10c..052b90fb17 100644 --- a/snapshots/session/ptc-read-image/system-prompt.expected.md +++ b/snapshots/session/ptc-read-image/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -420,7 +407,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/ptc-turn/system-prompt.expected.md b/snapshots/session/ptc-turn/system-prompt.expected.md index 1e964bdb4f..2801645046 100644 --- a/snapshots/session/ptc-turn/system-prompt.expected.md +++ b/snapshots/session/ptc-turn/system-prompt.expected.md @@ -140,13 +140,6 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ - list_subagent_models: { - /** Registered LLM provider id. */ - provider?: string; - /** Exact model id; requires provider. */ - model?: string; - } & Record; /** Read a UTF-8 text file and return line-numbered content. */ read: { /** Path to read, resolved by the filesystem backend. */ @@ -173,7 +166,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; @@ -181,12 +174,6 @@ interface ToolArgsMap { description: string; /** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */ prompt: string; - /** Child LLM provider id. */ - provider?: string; - /** Model id within that provider. */ - model?: string; - /** Model-specific reasoning effort id. */ - reasoning_effort?: string; } & Record; /** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn). Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. */ subagent_fork: { @@ -424,7 +411,6 @@ interface ToolOutputMap { parent?: string; depth?: number; })[]; - list_subagent_models: string; read: { /** Canonical absolute path in the filesystem execution world. */ path: string; diff --git a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json index 9127b4826a..95ffc0255b 100644 --- a/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json +++ b/snapshots/session/pty-tools-sandbox-backend/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json index c35e8cec15..d1a1729417 100644 --- a/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json +++ b/snapshots/session/pwsh-tool-turn/tool-schemas.expected.json @@ -187,23 +187,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "pwsh", "description": "Execute a PowerShell command (`pwsh -Command`) and return its stdout/stderr. Each call runs in a fresh pwsh process; pass `workdir` instead of using `cd`. Paths use native Windows form (`C:\\...`); read environment variables with `$env:NAME`. Managed `$env:DSH_*` variables expose current harness environment facts. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. On Windows a force-killed command settles as `[exit code: 1]` without a signal marker — treat it as an interruption, not a command failure. Provide `description` before `command` in the arguments. Before any delete or move, verify that the resolved absolute target path is the intended one; never run it against a computed path you have not checked. Do not assign to automatic variables such as `$HOME`; variable names are case-insensitive, so `$home` is the same read-only variable. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under mode]`, a policy denial: do not retry another way. Under the Windows sandbox, read-only pwsh runs in PowerShell ConstrainedLanguage mode, while workspace-write stays in FullLanguage unless host policy says otherwise. In read-only, prefer cmdlets and core types (`[string]`, `[datetime]`, `[regex]`, `[guid]`); .NET static calls (`[System.IO.*]::`, `[math]::`), `Add-Type`, COM objects, and reflection fail with \"only core types\" errors. `-f` formatting, property access, and core cmdlets work. In both confined modes, programs cannot open named pipes, so a command that captures another program's output through piped stdio (Node.js `child_process.spawn`/`exec` with the default `stdio: 'pipe'`) fails with EPERM, while `stdio: 'inherit'` and `stdio: 'ignore'` spawns work and PowerShell's own pipelines are unaffected. That EPERM is the documented boundary: do not retry the command another way — escalate the exact command once or restructure it to avoid capturing output.", @@ -312,7 +295,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -327,18 +310,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.1.expected.json b/snapshots/session/ralph-loop/tool-schemas.1.expected.json index d72f9a2ae8..02ebdc4baf 100644 --- a/snapshots/session/ralph-loop/tool-schemas.1.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "ralph", "description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.", @@ -419,7 +402,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -434,18 +417,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.2.expected.json b/snapshots/session/ralph-loop/tool-schemas.2.expected.json index d72f9a2ae8..02ebdc4baf 100644 --- a/snapshots/session/ralph-loop/tool-schemas.2.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.2.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "ralph", "description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.", @@ -419,7 +402,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -434,18 +417,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/ralph-loop/tool-schemas.expected.json b/snapshots/session/ralph-loop/tool-schemas.expected.json index 9fba724dcc..deabaf4a0d 100644 --- a/snapshots/session/ralph-loop/tool-schemas.expected.json +++ b/snapshots/session/ralph-loop/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "ralph", "description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.", @@ -376,7 +359,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -391,18 +374,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/session-query-spill/tool-schemas.expected.json b/snapshots/session/session-query-spill/tool-schemas.expected.json index 1c778ec2c9..9d66381297 100644 --- a/snapshots/session/session-query-spill/tool-schemas.expected.json +++ b/snapshots/session/session-query-spill/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -560,7 +543,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -575,18 +558,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json index 48a1ac30ce..46abe32771 100644 --- a/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json +++ b/snapshots/session/subagent-acp-diagnostic/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json index b4fa90e046..b8dc7c20a8 100644 --- a/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json +++ b/snapshots/session/subagent-child-question-rejection/tool-schemas.expected.json @@ -323,23 +323,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -419,7 +402,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -434,18 +417,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml b/snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml deleted file mode 100644 index 3b8eed677f..0000000000 --- a/snapshots/session/subagent-model-discovery-errors/runtime.cordis.yml +++ /dev/null @@ -1,3 +0,0 @@ -- insert: - - id: discovery-fixture - name: '../../../packages/subagent/tool-subagent/tests/fixtures/model-discovery.mjs' diff --git a/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl b/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl deleted file mode 100644 index 1a781e4739..0000000000 --- a/snapshots/session/subagent-model-discovery-errors/session.v4.jsonl +++ /dev/null @@ -1,23 +0,0 @@ -{"type":"session","version":4,"id":"{{session:1}}","createdAt":1783352044766,"cwd":"{{cwd}}","isSeeded":false,"delegationDepth":0} -{"type":"permission/preset","data":{"preset":"danger-full-access"}} -{"type":"sandbox/mode","data":{"mode":"danger-full-access"}} -{"type":"approval/policy","data":{"policy":"never"}} -{"type":"agent/inbox/spliced","data":{"target":"next-turn","start":0,"inserted":[{"content":[{"type":"text","text":"List the models available for subagents from discovery-fixture, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"}]}} -{"type":"turn/start","data":{"turn":1}} -{"type":"agent/inbox/spliced","data":{"target":"next-turn","start":0,"removedCount":1,"inserted":[]}} -{"type":"subagent/model-selection-policy","data":{"allowedModels":[{"provider":"discovery-fixture","model":"removed"},{"provider":"discovery-fixture","model":"healthy"}]}} -{"type":"step/start","data":{"turn":1,"step":1}} -{"type":"system/message","data":{"turn":1,"step":1,"message":{"role":"system","content":[{"type":"text","text":"{{system}}"}],"source":{"kind":"system-prompt"},"id":"{{message:2}}"}},"surfaceOp":"append"} -{"type":"user/message","data":{"content":[{"type":"text","text":"List the models available for subagents from discovery-fixture, then reply DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:1}}"},"surfaceOp":"append"} -{"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: danger-full-access. The DSH file sandbox does not restrict file modifications by available operations."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."}]},"role":"user","id":"{{message:3}}"},"surfaceOp":"append"} -{"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":"{{tools}}"},"reason":"initial"}} -{"type":"request/context","data":{"provider":"deepseek-official","model":"deepseek-v4-flash"}} -{"type":"session/title","data":{"title":"List the models available for","messageSeqs":[9],"source":{"kind":"fallback"}}} -{"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:4}}"},"usage":{"inputTokens":100,"outputTokens":10},"stream":[{"type":"chunk","time":1791788384782,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"chunk","time":1791788384782,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}}},{"type":"chunk","time":1791788384782,"chunk":{"type":"usage","usage":{"inputTokens":100,"outputTokens":10}}},{"type":"chunk","time":1791788384782,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} -{"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_discovery","name":"list_subagent_models","arguments":"{\"provider\":\"discovery-fixture\"}"}} -{"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_discovery"},"toolCallId":"call_discovery","content":[{"type":"text","text":"discovery-fixture/healthy — Healthy"}],"isError":false,"id":"{{message:5}}"}},"sourceEventSeqs":[15],"surfaceOp":"append"} -{"type":"step/end","data":{"turn":1,"step":1}} -{"type":"step/start","data":{"turn":1,"step":2}} -{"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:6}}"},"usage":{"inputTokens":100,"outputTokens":10},"stream":[{"type":"chunk","time":1791788384799,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"chunk","time":1791788384800,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1791788384800,"chunk":{"type":"usage","usage":{"inputTokens":100,"outputTokens":10}}},{"type":"chunk","time":1791788384800,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} -{"type":"step/end","data":{"turn":1,"step":2}} -{"type":"turn/end","data":{"turn":1,"reason":{"kind":"completed"}}} diff --git a/snapshots/session/subagent-model-discovery-errors/snapshot.yml b/snapshots/session/subagent-model-discovery-errors/snapshot.yml deleted file mode 100644 index 322ac5fd4f..0000000000 --- a/snapshots/session/subagent-model-discovery-errors/snapshot.yml +++ /dev/null @@ -1,7 +0,0 @@ -version: 1 -scenario: subagent-model-discovery-errors -profile: headless -composition: default -recording: authored -header: - class: default diff --git a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json index dd16436c73..e48983205c 100644 --- a/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-no-message-tool/tool-schemas.1.expected.json @@ -174,23 +174,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -249,7 +232,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -264,18 +247,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json index 16a256b818..d217c78441 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.1.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -371,7 +354,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -386,18 +369,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json index 498673be8a..25793c1cd4 100644 --- a/snapshots/session/subagent-structured-parent/tool-schemas.expected.json +++ b/snapshots/session/subagent-structured-parent/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json index 4442bcf828..45ab57c7c0 100644 --- a/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json +++ b/snapshots/session/subagent-tool-filter/tool-schemas.1.expected.json @@ -174,23 +174,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -270,7 +253,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -285,18 +268,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/system-prompt-in-history/tool-schemas.expected.json +++ b/snapshots/session/system-prompt-in-history/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/team-targets/tool-schemas.expected.json b/snapshots/session/team-targets/tool-schemas.expected.json index f319d5da9f..4926edb120 100644 --- a/snapshots/session/team-targets/tool-schemas.expected.json +++ b/snapshots/session/team-targets/tool-schemas.expected.json @@ -251,23 +251,6 @@ "properties": {} } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", diff --git a/snapshots/session/text-turn/tool-schemas.expected.json b/snapshots/session/text-turn/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/text-turn/tool-schemas.expected.json +++ b/snapshots/session/text-turn/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/web-fetch/tool-schemas.expected.json b/snapshots/session/web-fetch/tool-schemas.expected.json index 114d460759..5ef902c18a 100644 --- a/snapshots/session/web-fetch/tool-schemas.expected.json +++ b/snapshots/session/web-fetch/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/workflow-confinement/tool-schemas.expected.json b/snapshots/session/workflow-confinement/tool-schemas.expected.json index 185fd4ef83..2b4b1f069f 100644 --- a/snapshots/session/workflow-confinement/tool-schemas.expected.json +++ b/snapshots/session/workflow-confinement/tool-schemas.expected.json @@ -260,23 +260,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -356,7 +339,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -371,18 +354,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/workspace-dependencies/tool-schemas.expected.json b/snapshots/session/workspace-dependencies/tool-schemas.expected.json index f009027b9e..4d3db178b9 100644 --- a/snapshots/session/workspace-dependencies/tool-schemas.expected.json +++ b/snapshots/session/workspace-dependencies/tool-schemas.expected.json @@ -215,23 +215,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "load_workspace_dependencies", "description": "Get absolute paths to bundled Python and library directories, plus bundled Python distribution versions. Node.js and pnpm paths are included when the payload provides them. Python includes numpy, pandas, python-docx, python-pptx, openpyxl, Pillow, lxml, and XlsxWriter. Use these libraries for Office files unless the user or workspace instructions select another environment. When Node.js and pnpm paths are returned, run pnpm with that Node executable and pnpm script path. This does not change PATH or package-manager settings.", @@ -319,7 +302,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -334,18 +317,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/snapshots/session/worktree/tool-schemas.expected.json b/snapshots/session/worktree/tool-schemas.expected.json index d95e52e13e..6641f8f0b2 100644 --- a/snapshots/session/worktree/tool-schemas.expected.json +++ b/snapshots/session/worktree/tool-schemas.expected.json @@ -277,23 +277,6 @@ } } }, - { - "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", - "parameters": { - "type": "object", - "properties": { - "provider": { - "type": "string", - "description": "Registered LLM provider id." - }, - "model": { - "type": "string", - "description": "Exact model id; requires provider." - } - } - } - }, { "name": "read", "description": "Read a UTF-8 text file and return line-numbered content.", @@ -373,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.", "parameters": { "type": "object", "properties": { @@ -388,18 +371,6 @@ "prompt": { "type": "string", "description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs." - }, - "provider": { - "type": "string", - "description": "Child LLM provider id." - }, - "model": { - "type": "string", - "description": "Model id within that provider." - }, - "reasoning_effort": { - "type": "string", - "description": "Model-specific reasoning effort id." } }, "required": [ diff --git a/tsconfig.base.json b/tsconfig.base.json index dfb4faf2eb..0d5ec2c48a 100644 --- a/tsconfig.base.json +++ b/tsconfig.base.json @@ -185,7 +185,6 @@ "@deepseek-ai/dsh-tools/presentation": ["./packages/core/tools/src/presentation.ts"], "@deepseek-ai/dsh-tools/types": ["./packages/core/tools/src/types.ts"], "@deepseek-ai/dsh-tool-subagent-control/list-agents": ["./packages/subagent/tool-subagent-control/src/list-agents.ts"], - "@deepseek-ai/dsh-tool-subagent/list-models": ["./packages/subagent/tool-subagent/src/list-models.ts"], "@deepseek-ai/dsh-tool-subagent/model-selection-settings": ["./packages/subagent/tool-subagent/src/model-selection-settings.ts"], "@deepseek-ai/dsh-user-approval/types": ["./packages/interaction/user-approval/src/types.ts"], "@deepseek-ai/dsh-user-questions/types": ["./packages/interaction/user-questions/src/types.ts"], From 84c3a61abfb7724af22963dab095b6e5c82af068 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Tue, 13 Oct 2026 11:16:53 +0800 Subject: [PATCH 017/162] fix(subagent): keep fork model selection configurable --- ...authorized-subagent-model-routes.i18n.yaml | 4 +- ...4-user-authorized-subagent-model-routes.md | 2 +- ...ser-authorized-subagent-model-routes.zh.md | 2 +- docs/config-catalog.md | 4 +- docs/config-catalog.zh.md | 4 +- docs/tool-catalog.i18n.yaml | 8 ++-- docs/tool-catalog.md | 4 +- docs/tool-catalog.zh.md | 4 +- .../guide.i18n.yaml | 16 -------- .../fork-model-selection-settings/guide.md | 28 -------------- .../fork-model-selection-settings/guide.zh.md | 28 -------------- .../subagent/tool-subagent/README.i18n.yaml | 16 ++++---- packages/subagent/tool-subagent/README.md | 8 ++-- packages/subagent/tool-subagent/README.zh.md | 8 ++-- packages/subagent/tool-subagent/src/index.ts | 10 +---- .../tests/model-selection.spec.ts | 37 +++++++++++++++++-- scripts/gen-tool-catalog.ts | 2 +- 17 files changed, 68 insertions(+), 117 deletions(-) delete mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml delete mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md delete mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index 250e5e27cf..4561eacc4e 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -9,8 +9,8 @@ en: 3238dfb7db7dfdd8 zh: ec3bdd9a9da0a321 /agent-note-user-authorized-subagent-model-routes/decision: - en: e67644944167a865 - zh: dc2b28f630a0f6fa + en: 423eefc69964717c + zh: 7f73ea1c82630314 /agent-note-user-authorized-subagent-model-routes/alternatives-considered: en: e41c17226190ea8b zh: d3b1395fafe75f01 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index 357ef78e36..a44c5245e1 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -16,7 +16,7 @@ A newly composed top-level Session snapshots the route list in `subagent/model-s The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and intersects their advertised model catalogs with the recorded route list. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts, including for ids absent from the catalog. Without a recorded policy, discovery resolves only the parent Agent’s current effective route, even if the catalog omits it. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. -Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. Only a delegation tool configured with `modelSelectionSettings` exposes the route fields and registers `list_subagent_models`; the shipped Web spawn tool is that owner, captures the setting once, and reads the Session policy at each call. Fork tools reject model-selection arguments, and the setting fails their mount, to preserve inherited-prefix reuse. +Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. Only a delegation tool configured with `modelSelectionSettings` exposes the route fields and registers `list_subagent_models`; the shipped Web spawn tool is that owner, captures the setting once, and reads the Session policy at each call. Shipped fork tools leave `modelSelectionSettings` disabled to preserve inherited-prefix reuse; custom fork compositions may enable it under the same Session authorization. ## Alternatives considered diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index 4fe117478a..f5141f6c08 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -16,7 +16,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` 固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并取其公布模型目录与已记录路由列表的交集。精确查询先要求授权，再解析模型元数据及全部支持的推理强度，也允许目录未列出的 ID。没有已记录策略时，发现只解析父 Agent 当前生效的路由，即使目录未列出它。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 -模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。只有配置了 `modelSelectionSettings` 的委派工具会公开路由字段并注册 `list_subagent_models`；随附 Web 的 spawn 工具就是该所有者，它一次性记录设置，并在每次调用时读取 Session 策略。fork 工具拒绝模型选择参数，且该设置会使其挂载失败，以保留对继承前缀的复用。 +模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。只有配置了 `modelSelectionSettings` 的委派工具会公开路由字段并注册 `list_subagent_models`；随附 Web 的 spawn 工具就是该所有者，它一次性记录设置，并在每次调用时读取 Session 策略。随附 fork 工具保持 `modelSelectionSettings` 关闭，以保留对继承前缀的复用；自定义 fork 组合可以在相同 Session 授权下启用它。 ## Alternatives considered diff --git a/docs/config-catalog.md b/docs/config-catalog.md index 48f63cf574..fb0e67b20e 100644 --- a/docs/config-catalog.md +++ b/docs/config-catalog.md @@ -3993,8 +3993,8 @@ export interface Config { * Expose optional child `provider`, `model`, and `reasoning_effort` fields * plus `list_subagent_models`, and sample the Host `subagent-model-selection` * setting for each new top-level Session; child Sessions inherit that - * decision. Requires a provider with `agentOptions` that does not inherit - * the parent conversation. At most one instance per tool scope may enable it. + * decision. Requires a provider with `agentOptions` support. + * At most one instance per tool scope may enable it. */ modelSelectionSettings?: boolean /** diff --git a/docs/config-catalog.zh.md b/docs/config-catalog.zh.md index b462472c26..69942f48e7 100644 --- a/docs/config-catalog.zh.md +++ b/docs/config-catalog.zh.md @@ -3995,8 +3995,8 @@ export interface Config { * Expose optional child `provider`, `model`, and `reasoning_effort` fields * plus `list_subagent_models`, and sample the Host `subagent-model-selection` * setting for each new top-level Session; child Sessions inherit that - * decision. Requires a provider with `agentOptions` that does not inherit - * the parent conversation. At most one instance per tool scope may enable it. + * decision. Requires a provider with `agentOptions` support. + * At most one instance per tool scope may enable it. */ modelSelectionSettings?: boolean /** diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 529e9be439..3ca5b75f64 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: 551b31e8c713bbcd - zh: 861604a6d4d6db6b + en: 252109967b78c79a + zh: c1f807cee3f44f94 /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -228,8 +228,8 @@ en: 6872bee11fccb6c8 zh: 230f2d443dc5d9a6 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: d73236d49ab7c9f5 - zh: 1009b5c5d6d678a5 + en: c184808a51ea4000 + zh: 7a936839c719dea4 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 090afd911d..cf9c3ba93a 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. The shipped `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2067,7 +2067,7 @@ Delegate a self-contained task to a subagent (a separate agent that works in its Source: [`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. +The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. The shipped `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index b60ade9669..c94fd2a8fc 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。`subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。随附的 `subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2073,7 +2073,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， 来源：[`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。`subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 +注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。随附的 `subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml deleted file mode 100644 index c3a98c82ac..0000000000 --- a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.i18n.yaml +++ /dev/null @@ -1,16 +0,0 @@ -# Bilingual-pair consistency record for guide.md (docs/i18n/README.md): per heading -# section, a hash of its English and Chinese blocks outside code blocks and generated regions. -# After editing either side, bring the other along and re-record with: -# pnpm run verify-translation-pairing --write docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md -/: - en: 9a987818d6903fcb - zh: 8a4eefd0e7b4ad04 -/remove-modelselectionsettings-from-fork-delegation-tools: - en: f6bf4e381476a12f - zh: 0e745d803fc38fc9 -/remove-modelselectionsettings-from-fork-delegation-tools/change: - en: ccac82df3a00bf69 - zh: 1820a052adff5f6b -/remove-modelselectionsettings-from-fork-delegation-tools/migration: - en: ba409934bc81f492 - zh: fda2f67f731cd1d2 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md deleted file mode 100644 index 542cc9a183..0000000000 --- a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.md +++ /dev/null @@ -1,28 +0,0 @@ ---- -kind: upgrade-guide -description: "`modelSelectionSettings` on a fork delegation tool now fails the mount instead of exposing child route selection." ---- - -# Remove `modelSelectionSettings` from fork delegation tools - -English | [中文](guide.zh.md) - -## Change - -In v0.2.1-alpha.2, a `@deepseek-ai/dsh-tool-subagent` row whose provider inherits the parent conversation, such as the in-process `fork` provider, accepted `modelSelectionSettings: true` and exposed `provider`, `model`, `reasoning_effort`, and `list_subagent_models` in Sessions with an enabled policy. The tool now rejects that configuration when the provider is mounted, so the composition fails to load with `provider "" inherits the parent conversation and does not accept child model selection`. Fork children keep the parent route so the inherited conversation prefix stays eligible for KV Cache reuse. - -The shipped profiles are unaffected: only the Web spawn tool sets `modelSelectionSettings`. A tool row with the setting now exposes the route fields and `list_subagent_models` in every Session; without a recorded Session policy, explicit selection permits only the parent route. - -## Migration - -1. In `cordis.yml`, profile patches, and overlays, remove `modelSelectionSettings` from every `@deepseek-ai/dsh-tool-subagent` row whose `provider` names a fork provider: - - ```yaml - - name: '@deepseek-ai/dsh-tool-subagent' - config: - provider: fork - toolName: subagent_fork - ``` - -2. To let the model choose child routes, set `modelSelectionSettings: true` on one spawn-style row per tool scope instead. -3. Start the composition and confirm that it loads and that the fork tool schema lists only `cwd`, `description`, and `prompt`. diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md b/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md deleted file mode 100644 index e400299473..0000000000 --- a/docs/upgrade-guide/v0.2.1-alpha.2/fork-model-selection-settings/guide.zh.md +++ /dev/null @@ -1,28 +0,0 @@ ---- -kind: upgrade-guide -description: "fork 委派工具上的 `modelSelectionSettings` 现在会使挂载失败，而不再公开子级路由选择。" ---- - -# 从 fork 委派工具中移除 `modelSelectionSettings` - -[English](guide.md) | 中文 - -## 变更 - -在 v0.2.1-alpha.2 中，如果 `@deepseek-ai/dsh-tool-subagent` 配置行的提供方会继承父级对话（例如进程内 `fork` 提供方），它会接受 `modelSelectionSettings: true`，并在策略已启用的 Session 中公开 `provider`、`model`、`reasoning_effort` 和 `list_subagent_models`。现在该工具会在提供方挂载时拒绝这一配置，组合加载失败并报告 `provider "" inherits the parent conversation and does not accept child model selection`。fork 子级保持父级路由，使继承的对话前缀继续符合 KV Cache 复用条件。 - -随附 profile 不受影响：只有 Web 的 spawn 工具设置了 `modelSelectionSettings`。设置了该项的工具行现在会在每个 Session 中公开路由字段和 `list_subagent_models`；没有已记录 Session 策略时，显式选择只允许父级路由。 - -## 迁移 - -1. 在 `cordis.yml`、profile patch 和 overlay 中，从所有 `provider` 指向 fork 提供方的 `@deepseek-ai/dsh-tool-subagent` 配置行移除 `modelSelectionSettings`： - - ```yaml - - name: '@deepseek-ai/dsh-tool-subagent' - config: - provider: fork - toolName: subagent_fork - ``` - -2. 如需让模型选择子级路由，请改为在每个工具作用域中的一个 spawn 类配置行上设置 `modelSelectionSettings: true`。 -3. 启动组合，确认它能加载，且 fork 工具的 schema 只列出 `cwd`、`description` 和 `prompt`。 diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index a806e98777..f02f435211 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -18,14 +18,14 @@ en: 93f1701f5209ecb5 zh: 6310fd012400aead /deepseek-ai-dsh-tool-subagent/use-this-package/minimal-configuration: - en: 70e715a7262f45b2 - zh: a94e81d7f1513fdf + en: 5465759f3c7fd5d7 + zh: 3f0b0cba46dbbc08 /deepseek-ai-dsh-tool-subagent/use-this-package/managed-delegation: en: 42673064ba88a211 zh: 8fa26285bd649dc9 /deepseek-ai-dsh-tool-subagent/use-this-package/selecting-a-child-llm: - en: 2bf262d3f462f5f5 - zh: c443c5366f230eee + en: ff7357f058f309ba + zh: 489f05233c1785a7 /deepseek-ai-dsh-tool-subagent/understand-the-implementation: en: 6ce5f749256a7e6c zh: a4f2adb04a70d9cf @@ -60,8 +60,8 @@ en: 86d1dafcb6c00c57 zh: fe890e9c6146c45e /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/kv-cache-effect: - en: f746a3c2f2f3cd18 - zh: 68de857073b229ff + en: 5697c49cb84fb2a4 + zh: b2b8464d3cbabfa5 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery: en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 @@ -114,8 +114,8 @@ en: f5d5f97a153f7c70 zh: 3d6bc05b9eadfbc2 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work: - en: 8a0d48a82dff3343 - zh: 604ecd4e288ae322 + en: cbe53030678b9dd4 + zh: 06ccb6828e5708d5 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work/dev-note: en: e67f633dc6fe7773 zh: 3bb0c20c93f16f98 diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index 6f188766d6..b8de1764f4 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -46,7 +46,7 @@ Load the subagent service, an in-process or remote backend, and this tool; then |---|---|---| | `provider` | required | Provider name on `ctx.subagents` (e.g. `spawn`, `fork`, `acp`) | | `toolName` | `subagent` | Model-facing tool name; distinct for every loaded instance | -| `modelSelectionSettings` | `false` | Expose child route fields and `list_subagent_models`, and sample the Host's exact-route authorization preference for each top-level Session; a standing preset observes matching Sessions, while direct Agent setup passes its Session explicitly; requires provider `agentOptions` support without parent-context inheritance | +| `modelSelectionSettings` | `false` | Expose child route fields and `list_subagent_models`, and sample the Host's exact-route authorization preference for each top-level Session; a standing preset observes matching Sessions, while direct Agent setup passes its Session explicitly; requires provider `agentOptions` support | | `agentOptions` | — | Configured child `provider`, `model`, adapter-owned `reasoningEffort`, and positive `maxTokens` defaults; requires provider `agentOptions` support and overlays any provider-owned route defaults | | `persona` | — | Per-child persona; requires the provider's `persona` capability | | `toolFilter` | — | Per-child global-tool restriction; requires the `toolFilter` capability | @@ -64,7 +64,7 @@ Every backend’s completion notice carries its final answer. Local Spawn and Fo ### Selecting a child LLM -Set `modelSelectionSettings: true` to expose optional `provider`, `model`, and `reasoning_effort` fields and register `list_subagent_models` with this instance; instances without it expose neither and reject selection fields. The instance samples the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. This mode requires a backend that advertises `agentOptions` and does not inherit the parent conversation; both in-process Spawn and DSH SDK qualify, while Fork, ACP, Codex, and Claude Code fail the mount. +Set `modelSelectionSettings: true` to expose optional `provider`, `model`, and `reasoning_effort` fields and register `list_subagent_models` with this instance; instances without it expose neither and reject selection fields. The instance samples the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. This mode requires a backend that advertises `agentOptions`; in-process Spawn and Fork and DSH SDK qualify, while ACP, Codex, and Claude Code fail the mount. A call supplies `provider` and `model` together, or supplies only an effort when configured, parent, or provider-owned defaults provide the route. Static `provider.agentRouteDefaults`, when present, form the provider/model baseline; tool configuration and model fields overlay it before route-aware effort merging and exact-route preflight. Providers without these defaults use compatible values from the parent's latest logged request, then the parent's creation options before its first request, while retaining the configured `maxTokens`. Changing the route without an explicit effort clears the inherited route-owned effort, so the selected model resolves its default. The live LLM adapter validates the effective route before child creation. Catalog membership remains advisory, so a model can use an unlisted id when its adapter accepts it. @@ -136,7 +136,7 @@ Fixed schema cost per parent request; `modelSelectionSettings` adds three option #### KV Cache effect -Prefix-stable while provider instances and their configuration are unchanged. Adapter catalog changes do not alter the definition. Fork tools omit model-selection fields to preserve reuse of the inherited parent prefix. +Prefix-stable while provider instances and their configuration are unchanged. Adapter catalog changes do not alter the definition. Shipped fork tools omit model-selection fields to preserve reuse of the inherited parent prefix. A custom fork tool may enable selection; changing its route can prevent prefix reuse. ### Model selection and discovery @@ -208,7 +208,7 @@ Append-only; newly visible content follows the reusable request prefix and does These limits define what this tool does not return or enforce; they are current package constraints. - **This tool does not wait for results** — use the completion notice or child messages. Programmatic result waiting belongs to the subagent service. -- **Fork tools reject model selection** — tools that inherit parent context neither expose nor accept `provider`, `model`, or `reasoning_effort`, even with Session authorization, and `modelSelectionSettings` fails their mount. Configured `agentOptions` remain available to the deployment. +- **Shipped fork tools omit model selection** — their configuration leaves `modelSelectionSettings` disabled to preserve inherited-prefix reuse. Custom fork tools may enable it under the same Session authorization and adapter validation as other delegation tools. - **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection requires `modelSelectionSettings` and is restricted to the Session’s authorized routes, or the current parent route without an enabled policy. diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index 7a819691ca..575b9b4530 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -46,7 +46,7 @@ kind: "package-reference" |---|---|---| | `provider` | 必填 | `ctx.subagents` 上的提供方名称（如 `spawn`、`fork`、`acp`） | | `toolName` | `subagent` | 面向模型的工具名称；每个已加载实例必须不同 | -| `modelSelectionSettings` | `false` | 公开子级路由字段和 `list_subagent_models`，并为每个顶层 Session 读取宿主的精确路由授权偏好；常驻 preset 观察匹配 Session，直接 Agent setup 则显式传入其 Session；要求提供方支持 `agentOptions` 且不继承父级上下文 | +| `modelSelectionSettings` | `false` | 公开子级路由字段和 `list_subagent_models`，并为每个顶层 Session 读取宿主的精确路由授权偏好；常驻 preset 观察匹配 Session，直接 Agent setup 则显式传入其 Session；要求提供方支持 `agentOptions` | | `agentOptions` | — | 配置的子级 `provider`、`model`、适配器所有的 `reasoningEffort` 与正整数 `maxTokens` 默认值；要求提供方支持 `agentOptions`，并会覆盖提供方持有的路由默认值 | | `persona` | — | 每个子 agent 独立的 persona；要求提供方具备 `persona` 能力 | | `toolFilter` | — | 每个子 agent 独立的全局工具限制；要求提供方具备 `toolFilter` 能力 | @@ -64,7 +64,7 @@ kind: "package-reference" ### 选择子级 LLM -设置 `modelSelectionSettings: true` 后，该实例公开可选的 `provider`、`model` 和 `reasoning_effort` 字段，并注册 `list_subagent_models`；未设置的实例两者都不公开，并拒绝选择字段。该实例为新顶层 Session 读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。此模式要求后端公布 `agentOptions` 且不继承父级对话；进程内 Spawn 和 DSH SDK 满足条件，Fork、ACP、Codex 和 Claude Code 会使挂载失败。 +设置 `modelSelectionSettings: true` 后，该实例公开可选的 `provider`、`model` 和 `reasoning_effort` 字段，并注册 `list_subagent_models`；未设置的实例两者都不公开，并拒绝选择字段。该实例为新顶层 Session 读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。此模式要求后端公布 `agentOptions`；进程内 Spawn、Fork 和 DSH SDK 满足条件，ACP、Codex 和 Claude Code 会使挂载失败。 一次调用需同时提供 `provider` 与 `model`；当配置值、父 agent 值或提供方持有的默认值能提供路由时，也可只提供推理等级。静态的 `provider.agentRouteDefaults` 在存在时构成提供方／模型基线；工具配置与模型字段会在路由相关强度合并和确切路由预检前覆盖它。没有这些默认值的提供方会使用父 agent 最新已记录请求中的兼容值，再使用父级首次请求前的创建选项，并保留配置的 `maxTokens`。更改路由但未显式提供推理等级时，会清除继承的路由自有等级，使所选模型解析自己的默认值。实时 LLM 适配器在创建子 agent 前校验有效路由。目录成员资格只提供建议，因此适配器接受时，模型可以使用未列出的 id。 @@ -136,7 +136,7 @@ kind: "package-reference" #### KV Cache 影响 -只要提供方实例及其配置不变，前缀就保持稳定。适配器目录变化不会改变定义。fork 工具省略模型选择字段，以保留对继承的父级前缀的复用。 +只要提供方实例及其配置不变，前缀就保持稳定。适配器目录变化不会改变定义。随附 fork 工具省略模型选择字段，以保留对继承的父级前缀的复用。自定义 fork 工具可以启用选择；更改其路由可能使前缀无法复用。 ### 模型选择与发现 @@ -208,7 +208,7 @@ Start independent delegations with `subagent` or `subagent_fork` together in one 这些限制说明本工具不返回或不强制执行什么；它们是当前包约束。 - **本工具不等待结果**——使用完成通知或子任务消息。程序化结果等待由 subagent 服务提供。 -- **fork 工具拒绝模型选择** — 继承父级上下文的工具既不公开也不接受 `provider`、`model` 或 `reasoning_effort`，即使 Session 已授权也不例外；`modelSelectionSettings` 会使其挂载失败。部署仍可配置 `agentOptions`。 +- **随附 fork 工具省略模型选择**：其配置保持 `modelSelectionSettings` 关闭，以保留对继承前缀的复用。自定义 fork 工具可以启用它，并接受与其他委派工具相同的 Session 授权和适配器校验。 - **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择需要 `modelSelectionSettings`，并受 Session 授权路由限制，没有启用策略时只允许当前父级路由。 diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index 928e405f9f..d07e4aa883 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -57,8 +57,8 @@ export interface Config { * Expose optional child `provider`, `model`, and `reasoning_effort` fields * plus `list_subagent_models`, and sample the Host `subagent-model-selection` * setting for each new top-level Session; child Sessions inherit that - * decision. Requires a provider with `agentOptions` that does not inherit - * the parent conversation. At most one instance per tool scope may enable it. + * decision. Requires a provider with `agentOptions` support. + * At most one instance per tool scope may enable it. */ modelSelectionSettings?: boolean /** @@ -193,12 +193,6 @@ export function apply(ctx: Context, config: Config, session?: Session): void { `tool-subagent: provider "${subagentProvider.name}" does not support child model selection`, ) } - if (modelSelectionEnabled && subagentProvider.inheritsParentContext) { - throw new Error( - `tool-subagent: provider "${subagentProvider.name}" inherits the parent conversation and does not accept child model selection` - + ' — remove `modelSelectionSettings` from this instance', - ) - } } // Validate provider-owned config outside the optional LLM binding so an diff --git a/packages/subagent/tool-subagent/tests/model-selection.spec.ts b/packages/subagent/tool-subagent/tests/model-selection.spec.ts index 69b72ae876..0181e77b9c 100644 --- a/packages/subagent/tool-subagent/tests/model-selection.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection.spec.ts @@ -495,12 +495,41 @@ describe('dsh-tool-subagent model selection', () => { expect(resolve).not.toHaveBeenCalled() }) - it('rejects modelSelectionSettings on a provider that inherits the parent conversation', async () => { - await expect(setup({ provider: 'mock', withModelSelection: true }, { inheritsParentContext: true })) - .rejects.toThrow('provider "mock" inherits the parent conversation and does not accept child model selection') + it('enforces Session authorization when a fork tool explicitly enables model selection', async () => { + const requests: SubagentStartRequest[] = [] + const ctx = await setup({ provider: 'mock', withModelSelection: true }, { + inheritsParentContext: true, + onStart: (request) => { requests.push(request) }, + }) + ctx.llm.registerAdapter(['alpha'], new MockAdapter([], REASONING)) + const agent = modelSelectionSetupAgent(ctx) + const schema = ctx.tools.schemas(agent).find(entry => entry.name === 'subagent')! + expect(schema.description).toContain('inherits this conversation') + expect(schema.description).toContain('list_subagent_models') + + const allowed = await callSubagent(ctx, { + description: 'fork work', prompt: 'do it', provider: 'alpha', model: 'fast-model', reasoning_effort: 'low', + }) + expect(allowed.isError, text(allowed)).toBe(false) + expect(requests[0]?.agentOptions).toEqual({ + provider: 'alpha', model: 'fast-model', reasoningEffort: ReasoningEffortId('low'), + }) + + const denied = await callSubagent(ctx, { + description: 'unauthorized route', prompt: 'do it', provider: 'alpha', model: 'denied-model', + }) + expect(denied.isError).toBe(true) + expect(text(denied)).toContain('is not allowed for this Session') + + const invalidEffort = await callSubagent(ctx, { + description: 'unsupported effort', prompt: 'do it', provider: 'alpha', model: 'fast-model', reasoning_effort: 'invalid', + }) + expect(invalidEffort.isError).toBe(true) + expect(text(invalidEffort)).toContain('invalid') + expect(requests).toHaveLength(1) }) - it('keeps fork model selection unavailable even with Session authorization', async () => { + it('keeps fork model selection unavailable without instance opt-in even with Session authorization', async () => { const started = vi.fn() const ctx = await setup({ provider: 'mock' }, { inheritsParentContext: true, diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index 4cfa955c2d..2d554d3542 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -522,7 +522,7 @@ const TOOL_PACKAGES: ToolPackage[] = [ registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'mock', model: 'mock' }] })) }, note: - 'The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', + 'The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. The shipped `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', }, { pkg: '@deepseek-ai/dsh-tool-subagent-control', From 8fa8e9167db1f71ab66f2ebc495bdc3399467ee9 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Tue, 13 Oct 2026 11:42:31 +0800 Subject: [PATCH 018/162] refactor(subagent): separate model discovery and session authorization --- ...authorized-subagent-model-routes.i18n.yaml | 8 +- ...4-user-authorized-subagent-model-routes.md | 6 +- ...ser-authorized-subagent-model-routes.zh.md | 6 +- docs/config-catalog.md | 12 +- docs/config-catalog.zh.md | 12 +- docs/event-producer-consumer.md | 2 +- docs/event-producer-consumer.zh.md | 2 +- docs/subsystems/subagent.md | 4 +- docs/subsystems/subagent.zh.md | 4 +- docs/tool-catalog.i18n.yaml | 8 +- docs/tool-catalog.md | 4 +- docs/tool-catalog.zh.md | 4 +- .../guide.i18n.yaml | 16 + .../guide.md | 19 + .../guide.zh.md | 19 + packages/bundle/web-app/cordis.patch.yml | 4 +- .../bundle/web-app/presets/cordis.patch.yml | 4 +- packages/bundle/web-app/presets/ptc.patch.yml | 4 +- .../bundle/web-app/presets/standard.patch.yml | 4 +- .../tests/serial-listener-review.spec.ts | 13 +- .../extensions/tool-cordis/src/api-catalog.ts | 6 +- .../fixtures/loader/scoped-tool-subagent.ts | 6 +- .../subagent/tool-subagent/README.i18n.yaml | 32 +- packages/subagent/tool-subagent/README.md | 18 +- packages/subagent/tool-subagent/README.zh.md | 18 +- packages/subagent/tool-subagent/package.json | 7 +- packages/subagent/tool-subagent/src/index.ts | 403 +++++++----------- .../subagent/tool-subagent/src/list-models.ts | 20 +- .../src/model-selection-settings.ts | 35 +- .../src/model-selection-state.ts | 17 - .../subagent/tool-subagent/tests/harness.ts | 9 +- .../tool-subagent/tests/list-models.spec.ts | 98 ++--- .../tests/model-selection-settings.spec.ts | 109 ++--- .../tests/model-selection.spec.ts | 2 +- .../subagent/tool-subagent/tsdown.config.ts | 1 + scripts/check-workspace-constraints.ts | 2 +- scripts/gen-tool-catalog.ts | 6 +- tsconfig.base.json | 1 + 38 files changed, 434 insertions(+), 511 deletions(-) create mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml create mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md create mode 100644 docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml index 4561eacc4e..08799efa2b 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.i18n.yaml @@ -9,14 +9,14 @@ en: 3238dfb7db7dfdd8 zh: ec3bdd9a9da0a321 /agent-note-user-authorized-subagent-model-routes/decision: - en: 423eefc69964717c - zh: 7f73ea1c82630314 + en: 78824ae164ea974e + zh: 84aa7740d9790563 /agent-note-user-authorized-subagent-model-routes/alternatives-considered: en: e41c17226190ea8b zh: d3b1395fafe75f01 /agent-note-user-authorized-subagent-model-routes/consequences: - en: 36b76692f1605e6d - zh: a63a249c980bc654 + en: 77677b5828d0de29 + zh: 609109e3388933a9 /agent-note-user-authorized-subagent-model-routes/related-decisions: en: 0eead87ec1172d81 zh: 4e776a583f5a00a5 diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md index a44c5245e1..8113ddef6b 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.md @@ -12,11 +12,11 @@ Registering an LLM adapter makes its routes reachable, but does not authorize an The Host-owned `subagent-model-selection` settings section stores an explicit `enabled` switch and `allowedModels`, an array of exact `{ provider, model }` routes. Enabling requires at least one route; disabling may retain the selected routes for later reuse. The Plugins settings card reads the live adapter directory through `session/modelCatalog`, lets the user stage the switch and routes, and saves both fields in one revision-fenced settings mutation. It stores no adapter-owned display names, descriptions, or reasoning-effort metadata. A stored or staged route absent from the current directory remains visible as unavailable and removable; a provider-local catalog failure does not block other providers or erase saved authorization or an unsaved selection. A connection reset discards the draft because namespace revisions are comparable only within one Host process. -A newly composed top-level Session snapshots the route list in `subagent/model-selection-policy` when the setting is enabled, before its first model request. Event presence means selection was enabled; the event does not store the global switch. Child Sessions inherit that exact list from their live parent, and resumed Sessions use the recorded event instead of current settings. Settings changes therefore affect only subsequently composed top-level Sessions. Without the event, including an explicitly empty restored Session, explicit selection remains restricted to the current parent route. +The Host settings service snapshots the route list for each fresh top-level Session in `subagent/model-selection-policy` when the setting is enabled, before its first model request. Event presence means selection was enabled; the event does not store the global switch. Child Sessions inherit that exact list from their live parent, and resumed Sessions use the recorded event instead of current settings. Settings changes therefore affect only subsequently created top-level Sessions. Without the event, including an explicitly empty restored Session, explicit selection remains restricted to the current parent route. The fixed `list_subagent_models` schema does not enumerate the policy. At call time, it lists authorized registered providers and intersects their advertised model catalogs with the recorded route list. Exact lookup requires authorization before resolving model metadata and all supported reasoning efforts, including for ids absent from the catalog. Without a recorded policy, discovery resolves only the parent Agent’s current effective route, even if the catalog omits it. The delegation executor independently checks any explicit provider, model, or effort selection against the same authorization before adapter preflight. A call with no selection fields retains configured or inherited routing. -Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. Only a delegation tool configured with `modelSelectionSettings` exposes the route fields and registers `list_subagent_models`; the shipped Web spawn tool is that owner, captures the setting once, and reads the Session policy at each call. Shipped fork tools leave `modelSelectionSettings` disabled to preserve inherited-prefix reuse; custom fork compositions may enable it under the same Session authorization. +Model selection has no unrestricted static mode. The default-off Host setting authorizes alternative child routes; without it, the parent route and its supported efforts remain selectable. Each delegation tool’s `modelSelection` option controls only its route fields. The independent `/list-models` plugin owns discovery. Web and Desktop presets mount discovery by default, enable spawn route fields, and leave fork route fields disabled to preserve inherited-prefix reuse; custom fork compositions may enable them under the same Session authorization. Tool installation does not sample or expand authorization. ## Alternatives considered @@ -34,7 +34,7 @@ Model selection has no unrestricted static mode. The default-off Host setting au - New adapter registrations and newly advertised models do not expand user authorization. - Adapter removal or catalog lookup failure can make provider discovery unavailable without deleting the saved route decision; an allowed model remains usable when its adapter accepts it even if the advisory catalog omits it. -- The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. Compositions with `modelSelectionSettings` carry the three route fields and the discovery schema in every Session; headless and SDK profiles carry neither. +- The allowlist itself consumes no parent-request tokens. Only a `list_subagent_models` result enters the transcript. The `modelSelection` option adds three route fields; `/list-models` adds a separate discovery schema. Shipped headless and SDK profiles carry neither. - The policy event is log-only and is appended while an Agent is composed, before either SDK begins its run subscription. Shipped SDK profiles do not enable this Web-owned preference, so the event changes neither SDK's expected notifications or persisted-session output; package restore tests own its durable projection instead of fabricating an SDK composition solely to emit it. - Unit coverage pins settings validation, malformed durable values, Session sampling and inheritance, shared definitions, parent-route defaults, discovery restrictions, executor denial, and the settings UI’s staged writes and revision checks. The assembled Web scenario pins the real settings document and Plugins card flow. diff --git a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md index f5141f6c08..6d33817449 100644 --- a/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md +++ b/.agents/notes/implemented/feature/2026-08-24-user-authorized-subagent-model-routes.zh.md @@ -12,11 +12,11 @@ Status: implemented Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` 开关与 `allowedModels`，后者是由精确 `{ provider, model }` 路由组成的数组。启用时必须至少有一条路由；关闭时可以保留已选路由，供以后重新启用。Plugins 设置卡通过 `session/modelCatalog` 读取实时适配器目录，让用户暂存开关与路由，再在一次带 revision 限制的设置 mutation 中保存两个字段。它不保存适配器自有的显示名称、描述或推理强度元数据。当前目录中缺失的已存或暂存路由仍显示为不可用并允许移除；某个提供方的目录失败不会阻塞其他提供方，也不会清除已存授权或未保存选择。连接重置会丢弃草稿，因为 namespace revision 只能在同一个 Host 进程内比较。 -设置启用时，新组合的顶层 Session 在首个模型请求前将路由列表记录为 `subagent/model-selection-policy`。事件存在表示选择已启用；事件不保存全局开关。子 Session 从在线父级继承同一精确列表，恢复的 Session 使用已记录事件而非当前设置。因此设置变更只影响之后组合的顶层 Session。没有该事件时，包括显式为空的恢复 Session，显式选择仍仅限当前父级路由。 +设置启用时，Host 设置服务为每个新顶层 Session 在首个模型请求前将路由列表记录为 `subagent/model-selection-policy`。事件存在表示选择已启用；事件不保存全局开关。子 Session 从在线父级继承同一精确列表，恢复的 Session 使用已记录事件而非当前设置。因此设置变更只影响之后创建的顶层 Session。没有该事件时，包括显式为空的恢复 Session，显式选择仍仅限当前父级路由。 固定的 `list_subagent_models` schema 不枚举策略。调用时，它列出已授权且已注册的提供方，并取其公布模型目录与已记录路由列表的交集。精确查询先要求授权，再解析模型元数据及全部支持的推理强度，也允许目录未列出的 ID。没有已记录策略时，发现只解析父 Agent 当前生效的路由，即使目录未列出它。委派执行器会在适配器预检前独立校验显式提供方、模型或强度选择是否符合相同授权。没有选择字段的调用保留配置或继承的路由。 -模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。只有配置了 `modelSelectionSettings` 的委派工具会公开路由字段并注册 `list_subagent_models`；随附 Web 的 spawn 工具就是该所有者，它一次性记录设置，并在每次调用时读取 Session 策略。随附 fork 工具保持 `modelSelectionSettings` 关闭，以保留对继承前缀的复用；自定义 fork 组合可以在相同 Session 授权下启用它。 +模型选择没有无限制的静态模式。默认关闭的 Host 设置授权其他子级路由；没有该授权时，仍可选择父级路由及其支持的强度。各委派工具的 `modelSelection` 选项只控制自身路由字段。独立的 `/list-models` 插件负责发现。Web 与 Desktop preset 默认挂载发现工具，启用 spawn 路由字段，并为 fork 保持关闭，以保留对继承前缀的复用；自定义 fork 组合可以在相同 Session 授权下启用这些字段。工具安装不采样或扩大授权。 ## Alternatives considered @@ -34,7 +34,7 @@ Host 自有的 `subagent-model-selection` 设置 section 保存显式 `enabled` - 新适配器注册和新公布模型不会扩大用户授权。 - 适配器移除或目录查询失败可能使提供方发现不可用，但不会删除已存路由决定；只要适配器接受某个获准模型，即使建议性目录省略它，该模型仍可使用。 -- 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。配置了 `modelSelectionSettings` 的组合在每个 Session 中都携带三个路由字段和发现 schema；headless 与 SDK profile 两者都不携带。 +- 允许列表本身不消耗父级请求 token。只有 `list_subagent_models` 结果进入 transcript。`modelSelection` 选项增加三个路由字段；`/list-models` 增加独立发现 schema。随附 headless 与 SDK profile 两者都不携带。 - 策略事件仅存在于日志，并在 Agent 组合期间、两套 SDK 开始订阅运行前追加。随附 SDK profile 不启用这项 Web 自有偏好，因此该事件不会改变任一 SDK 的预期通知或持久 Session 输出；其持久投影由包级恢复测试负责，不会为了发出该事件而虚构 SDK 组合。 - 单元覆盖固定设置校验、异常持久值、Session 取样与继承、共享定义、父级路由默认值、发现限制、执行器拒绝，以及设置 UI 的暂存写入与 revision 检查。组装 Web 场景固定真实设置文档与 Plugins 设置卡流程。 diff --git a/docs/config-catalog.md b/docs/config-catalog.md index fb0e67b20e..76e25a9c29 100644 --- a/docs/config-catalog.md +++ b/docs/config-catalog.md @@ -3977,7 +3977,7 @@ export interface Config { - `inject`: `tools` · `subagents` · `systemPrompt` · `sessionProjections` - `refs`: [`AgentOptions`](subsystems/core.md) -- `source`: [`packages/subagent/tool-subagent/src/index.ts:48`](../packages/subagent/tool-subagent/src/index.ts) +- `source`: [`packages/subagent/tool-subagent/src/index.ts:42`](../packages/subagent/tool-subagent/src/index.ts) ```ts config-catalog /** Config: which registered provider this tool delegates to, plus child defaults. */ @@ -3990,13 +3990,11 @@ export interface Config { */ toolName?: string /** - * Expose optional child `provider`, `model`, and `reasoning_effort` fields - * plus `list_subagent_models`, and sample the Host `subagent-model-selection` - * setting for each new top-level Session; child Sessions inherit that - * decision. Requires a provider with `agentOptions` support. - * At most one instance per tool scope may enable it. + * Expose optional child `provider`, `model`, and `reasoning_effort` fields. + * Requires provider `agentOptions` support. Execution enforces the Session + * policy, or restricts explicit selection to the current parent route. */ - modelSelectionSettings?: boolean + modelSelection?: boolean /** * Agent options applied to every child; omitted fields use child-loop defaults. */ diff --git a/docs/config-catalog.zh.md b/docs/config-catalog.zh.md index 69942f48e7..e6b40be645 100644 --- a/docs/config-catalog.zh.md +++ b/docs/config-catalog.zh.md @@ -3979,7 +3979,7 @@ export interface Config { - `inject`: `tools` · `subagents` · `systemPrompt` · `sessionProjections` - `refs`: [`AgentOptions`](subsystems/core.zh.md) -- `source`: [`packages/subagent/tool-subagent/src/index.ts:48`](../packages/subagent/tool-subagent/src/index.ts) +- `source`: [`packages/subagent/tool-subagent/src/index.ts:42`](../packages/subagent/tool-subagent/src/index.ts) ```ts config-catalog /** Config: which registered provider this tool delegates to, plus child defaults. */ @@ -3992,13 +3992,11 @@ export interface Config { */ toolName?: string /** - * Expose optional child `provider`, `model`, and `reasoning_effort` fields - * plus `list_subagent_models`, and sample the Host `subagent-model-selection` - * setting for each new top-level Session; child Sessions inherit that - * decision. Requires a provider with `agentOptions` support. - * At most one instance per tool scope may enable it. + * Expose optional child `provider`, `model`, and `reasoning_effort` fields. + * Requires provider `agentOptions` support. Execution enforces the Session + * policy, or restricts explicit selection to the current parent route. */ - modelSelectionSettings?: boolean + modelSelection?: boolean /** * Agent options applied to every child; omitted fields use child-loop defaults. */ diff --git a/docs/event-producer-consumer.md b/docs/event-producer-consumer.md index b40bcf5462..360fa5bd1f 100644 --- a/docs/event-producer-consumer.md +++ b/docs/event-producer-consumer.md @@ -73,7 +73,7 @@ This matrix shows which packages dispatch each harness-owned event and which pac | `subagent/start` | `emit` | [`packages/subagent/subagent/src/index.ts:158`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`hooks-claude-code`](../packages/experimental/hooks-claude-code) | | `system-prompt/assemble` | `waterfall` | [`packages/core/system-prompt/src/index.ts:31`](../packages/core/system-prompt/src/index.ts) | [`system-prompt`](../packages/core/system-prompt) (`waterfall`) | [`agent`](../packages/core/agent), `browser-use-runtime`, [`session-reference`](../packages/context/session-reference), [`working-directory`](../packages/session/working-directory) | | `system-prompt/change` | `emit` | [`packages/core/system-prompt/src/index.ts:37`](../packages/core/system-prompt/src/index.ts) | [`system-prompt`](../packages/core/system-prompt) (`emit`) | - | -| `tools/change` | `emit` | [`packages/core/tools/src/index.ts:209`](../packages/core/tools/src/index.ts) | [`agent-preset-registry`](../packages/preset/agent-preset-registry) (`emit`), [`tools`](../packages/core/tools) (`emit`) | `browser-use-runtime`, [`tool-subagent`](../packages/subagent/tool-subagent) | +| `tools/change` | `emit` | [`packages/core/tools/src/index.ts:209`](../packages/core/tools/src/index.ts) | [`agent-preset-registry`](../packages/preset/agent-preset-registry) (`emit`), [`tools`](../packages/core/tools) (`emit`) | `browser-use-runtime` | | `tools/execute` | `waterfall` | [`packages/core/tools/src/index.ts:165`](../packages/core/tools/src/index.ts) | [`tools`](../packages/core/tools) (`waterfall`) | `browser-use-runtime`, `browser-use-stagehand-native`, `claude-code-mods`, `computer-use-cua-driver-native`, [`session-checkpoint-policy`](../packages/session/session-checkpoint-policy), `timeout-policy` | | `tools/post-execute` | `waterfall` | [`packages/core/tools/src/index.ts:177`](../packages/core/tools/src/index.ts) | [`tools`](../packages/core/tools) (`waterfall`) | `claude-code-mods`, [`hooks-claude-code`](../packages/experimental/hooks-claude-code), [`hooks-codex`](../packages/experimental/hooks-codex), [`repeat-tool-reminder`](../packages/guard/repeat-tool-reminder), [`spill-policy`](../packages/spill/spill-policy), [`tool-fs-search`](../packages/fs/tool-fs-search) | | `tools/pre-execute` | `waterfall` | [`packages/core/tools/src/index.ts:154`](../packages/core/tools/src/index.ts) | [`tools`](../packages/core/tools) (`waterfall`) | `auto-review`, [`hooks-claude-code`](../packages/experimental/hooks-claude-code), [`hooks-codex`](../packages/experimental/hooks-codex), [`tool-jobs`](../packages/jobs/tool-jobs), [`workspace-changes`](../packages/deliverables/workspace-changes) | diff --git a/docs/event-producer-consumer.zh.md b/docs/event-producer-consumer.zh.md index 9c7acd09bb..aa57743c7e 100644 --- a/docs/event-producer-consumer.zh.md +++ b/docs/event-producer-consumer.zh.md @@ -75,7 +75,7 @@ | `subagent/start` | `emit` | [`packages/subagent/subagent/src/index.ts:158`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`hooks-claude-code`](../packages/experimental/hooks-claude-code) | | `system-prompt/assemble` | `waterfall` | [`packages/core/system-prompt/src/index.ts:31`](../packages/core/system-prompt/src/index.ts) | [`system-prompt`](../packages/core/system-prompt) (`waterfall`) | [`agent`](../packages/core/agent), `browser-use-runtime`, [`session-reference`](../packages/context/session-reference), [`working-directory`](../packages/session/working-directory) | | `system-prompt/change` | `emit` | [`packages/core/system-prompt/src/index.ts:37`](../packages/core/system-prompt/src/index.ts) | [`system-prompt`](../packages/core/system-prompt) (`emit`) | - | -| `tools/change` | `emit` | [`packages/core/tools/src/index.ts:209`](../packages/core/tools/src/index.ts) | [`agent-preset-registry`](../packages/preset/agent-preset-registry) (`emit`), [`tools`](../packages/core/tools) (`emit`) | `browser-use-runtime`, [`tool-subagent`](../packages/subagent/tool-subagent) | +| `tools/change` | `emit` | [`packages/core/tools/src/index.ts:209`](../packages/core/tools/src/index.ts) | [`agent-preset-registry`](../packages/preset/agent-preset-registry) (`emit`), [`tools`](../packages/core/tools) (`emit`) | `browser-use-runtime` | | `tools/execute` | `waterfall` | [`packages/core/tools/src/index.ts:165`](../packages/core/tools/src/index.ts) | [`tools`](../packages/core/tools) (`waterfall`) | `browser-use-runtime`, `browser-use-stagehand-native`, `claude-code-mods`, `computer-use-cua-driver-native`, [`session-checkpoint-policy`](../packages/session/session-checkpoint-policy), `timeout-policy` | | `tools/post-execute` | `waterfall` | [`packages/core/tools/src/index.ts:177`](../packages/core/tools/src/index.ts) | [`tools`](../packages/core/tools) (`waterfall`) | `claude-code-mods`, [`hooks-claude-code`](../packages/experimental/hooks-claude-code), [`hooks-codex`](../packages/experimental/hooks-codex), [`repeat-tool-reminder`](../packages/guard/repeat-tool-reminder), [`spill-policy`](../packages/spill/spill-policy), [`tool-fs-search`](../packages/fs/tool-fs-search) | | `tools/pre-execute` | `waterfall` | [`packages/core/tools/src/index.ts:154`](../packages/core/tools/src/index.ts) | [`tools`](../packages/core/tools) (`waterfall`) | `auto-review`, [`hooks-claude-code`](../packages/experimental/hooks-claude-code), [`hooks-codex`](../packages/experimental/hooks-codex), [`tool-jobs`](../packages/jobs/tool-jobs), [`workspace-changes`](../packages/deliverables/workspace-changes) | diff --git a/docs/subsystems/subagent.md b/docs/subsystems/subagent.md index 5647fab452..31b95187e9 100644 --- a/docs/subsystems/subagent.md +++ b/docs/subsystems/subagent.md @@ -412,11 +412,11 @@ Generated from source by `scripts/gen-cordis-catalog.ts` (verified fresh by `pnp ### `ctx.subagentModelSelection` — `SubagentModelSelectionConfig` -Singleton settings owner read when delegation tools are composed for a Session. +Host settings owner that records Session authorization before the first model request. ```ts cordis-catalog /** - * Read a detached selection preference for the next eligible Session composition. + * Read a detached selection preference for the next fresh top-level Session. * @returns the enabled state and exact allowed routes. */ current(): SubagentModelSelectionSettings diff --git a/docs/subsystems/subagent.zh.md b/docs/subsystems/subagent.zh.md index 30262036b7..8733129909 100644 --- a/docs/subsystems/subagent.zh.md +++ b/docs/subsystems/subagent.zh.md @@ -414,11 +414,11 @@ Generated from source by `scripts/gen-cordis-catalog.ts` (verified fresh by `pnp ### `ctx.subagentModelSelection` — `SubagentModelSelectionConfig` -Singleton settings owner read when delegation tools are composed for a Session. +Host settings owner that records Session authorization before the first model request. ```ts cordis-catalog /** - * Read a detached selection preference for the next eligible Session composition. + * Read a detached selection preference for the next fresh top-level Session. * @returns the enabled state and exact allowed routes. */ current(): SubagentModelSelectionSettings diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 3ca5b75f64..86a316406d 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: 252109967b78c79a - zh: c1f807cee3f44f94 + en: 7224f159dd69abba + zh: 00b69ab3f899010a /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -228,8 +228,8 @@ en: 6872bee11fccb6c8 zh: 230f2d443dc5d9a6 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: c184808a51ea4000 - zh: 7a936839c719dea4 + en: 7b06bc34df756d56 + zh: bc82dab293f57b5b /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index cf9c3ba93a..7ff20028a7 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. The shipped `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave fork route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2067,7 +2067,7 @@ Delegate a self-contained task to a subagent (a separate agent that works in its Source: [`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. The shipped `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. +The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave fork route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index c94fd2a8fc..5231594677 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。随附的 `subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 fork 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2073,7 +2073,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， 来源：[`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -注册的委派工具名称取决于加载时 `toolName` 配置（默认为 `subagent`）；上述默认 schema 关闭模型选择，而发现 schema 展示为设置了 `modelSelectionSettings` 的实例注册的固定配套工具。该实例始终公开可选路由字段；Web preset 会在每个新顶层 Session 创建时读取插件页偏好，并为其子 Session 保留该决定；没有启用策略时，选择和发现仅限当前父级路由。随附的 `subagent_fork` 不公开路由选择。每个实例通过 `provider` 选择后端。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 +加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 fork 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml new file mode 100644 index 0000000000..5cc66aa405 --- /dev/null +++ b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml @@ -0,0 +1,16 @@ +# Bilingual-pair consistency record for guide.md (docs/i18n/README.md): per heading +# section, a hash of its English and Chinese blocks outside code blocks and generated regions. +# After editing either side, bring the other along and re-record with: +# pnpm run verify-translation-pairing --write docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md +/: + en: b9bbc14e49aa6bdf + zh: ce5f3d2b0d6a9ff9 +/separate-subagent-model-selection-configuration: + en: c54f686a434d8b84 + zh: f32b63140d9b2dda +/separate-subagent-model-selection-configuration/change: + en: 6773f124a0ff9f6a + zh: 99213e0ba2caff35 +/separate-subagent-model-selection-configuration/migration: + en: b83bd0256fb89c92 + zh: 5b14cf7cfa9677b3 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md new file mode 100644 index 0000000000..72411e4bc9 --- /dev/null +++ b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md @@ -0,0 +1,19 @@ +--- +kind: upgrade-guide +description: Custom subagent compositions must separate route fields, model discovery, and Session authorization. +--- + +# Separate subagent model selection configuration + +English | [中文](guide.zh.md) + +## Change + +`@deepseek-ai/dsh-tool-subagent` replaces `modelSelectionSettings` with `modelSelection`, which controls only optional child route fields. The tool no longer registers `list_subagent_models` or samples user settings. The Host `/model-selection-settings` service records authorization during Agent creation, including Sessions created before delegation tools are mounted. Existing Sessions keep their recorded policy; installing tools later does not resample settings. Shipped Web and Desktop presets include the independent discovery plugin and keep fork route fields disabled. Custom fork tools may enable them. + +## Migration + +1. In custom `cordis.yml`, patches, or direct plugin configuration, rename `modelSelectionSettings: true` to `modelSelection: true` on delegation tools that expose route arguments. +2. Add an independent `@deepseek-ai/dsh-tool-subagent/list-models` row to each tool scope that needs discovery. Multiple delegation tools may share it. +3. Mount `@deepseek-ai/dsh-tool-subagent/model-selection-settings` in the Host before creating Agents. Its required services are `sessions` and `sessionProjections`. Direct `apply(ctx, config)` callers no longer pass a Session argument. +4. Create a fresh Session and verify that discovery returns the authorized routes when the user preference is enabled, or only the parent Agent’s current route otherwise. Verify that an explicit unauthorized delegation fails even without first calling discovery. diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md new file mode 100644 index 0000000000..f312728181 --- /dev/null +++ b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md @@ -0,0 +1,19 @@ +--- +kind: upgrade-guide +description: 自定义 subagent 组合必须分离路由字段、模型发现和 Session 授权。 +--- + +# 分离 subagent 模型选择配置 + +[English](guide.md) | 中文 + +## 变更 + +`@deepseek-ai/dsh-tool-subagent` 将 `modelSelectionSettings` 替换为 `modelSelection`，该选项只控制可选子级路由字段。工具不再注册 `list_subagent_models` 或采样用户设置。Host 的 `/model-selection-settings` 服务在 Agent 创建时记录授权，包括在委派工具挂载前创建的 Session。已有 Session 保留已记录策略；之后安装工具不会重新采样设置。随附 Web 与 Desktop preset 包含独立发现插件，并为 fork 保持路由字段关闭。自定义 fork 工具可以启用这些字段。 + +## 迁移 + +1. 在自定义 `cordis.yml`、patch 或直接插件配置中，将公开路由参数的委派工具的 `modelSelectionSettings: true` 改为 `modelSelection: true`。 +2. 在每个需要发现的工具作用域中，添加独立的 `@deepseek-ai/dsh-tool-subagent/list-models` 配置行。多个委派工具可以共用它。 +3. 在创建 Agent 前，将 `@deepseek-ai/dsh-tool-subagent/model-selection-settings` 挂载到 Host。它需要 `sessions` 和 `sessionProjections` 服务。直接调用 `apply(ctx, config)` 时不再传入 Session 参数。 +4. 创建一个新 Session，确认用户偏好启用时发现结果为授权路由，否则仅为父 Agent 当前路由。确认即使未先调用发现工具，显式选择未授权路由的委派仍会失败。 diff --git a/packages/bundle/web-app/cordis.patch.yml b/packages/bundle/web-app/cordis.patch.yml index a1ca0b5483..540b7650da 100644 --- a/packages/bundle/web-app/cordis.patch.yml +++ b/packages/bundle/web-app/cordis.patch.yml @@ -61,8 +61,8 @@ enabled: true appVersion: !!js process.env.DSH_CLIENT_VERSION - # Host-owned opt-in sampled when a new Web session receives its preset - # delegation tools. The Plugins page edits this settings namespace. + # Host-owned authorization sampled at Agent creation. The Plugins page + # edits this settings namespace independently of the selected preset. - id: subagent-model-selection-settings name: '@deepseek-ai/dsh-tool-subagent/model-selection-settings' diff --git a/packages/bundle/web-app/presets/cordis.patch.yml b/packages/bundle/web-app/presets/cordis.patch.yml index 8c8d567ae3..e586ec9bad 100644 --- a/packages/bundle/web-app/presets/cordis.patch.yml +++ b/packages/bundle/web-app/presets/cordis.patch.yml @@ -89,12 +89,14 @@ name: '@deepseek-ai/dsh-tool-subagent-control' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn toolName: subagent - modelSelectionSettings: true + modelSelection: true toolFilter: deny: - schedule_create diff --git a/packages/bundle/web-app/presets/ptc.patch.yml b/packages/bundle/web-app/presets/ptc.patch.yml index 8e1ec1fb70..83f3b94b17 100644 --- a/packages/bundle/web-app/presets/ptc.patch.yml +++ b/packages/bundle/web-app/presets/ptc.patch.yml @@ -90,12 +90,14 @@ name: '@deepseek-ai/dsh-tool-subagent-control' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn toolName: subagent - modelSelectionSettings: true + modelSelection: true toolFilter: deny: - schedule_create diff --git a/packages/bundle/web-app/presets/standard.patch.yml b/packages/bundle/web-app/presets/standard.patch.yml index 6a42d0a351..ef64df96e9 100644 --- a/packages/bundle/web-app/presets/standard.patch.yml +++ b/packages/bundle/web-app/presets/standard.patch.yml @@ -90,12 +90,14 @@ name: '@deepseek-ai/dsh-tool-subagent-control' - id: tool-subagent-list-agents name: '@deepseek-ai/dsh-tool-subagent-control/list-agents' + - id: tool-subagent-list-models + name: '@deepseek-ai/dsh-tool-subagent/list-models' - id: tool-subagent name: '@deepseek-ai/dsh-tool-subagent' config: provider: spawn toolName: subagent - modelSelectionSettings: true + modelSelection: true toolFilter: deny: - schedule_create diff --git a/packages/core/agent-loop/tests/serial-listener-review.spec.ts b/packages/core/agent-loop/tests/serial-listener-review.spec.ts index fcd9f0aabd..818b36520e 100644 --- a/packages/core/agent-loop/tests/serial-listener-review.spec.ts +++ b/packages/core/agent-loop/tests/serial-listener-review.spec.ts @@ -8,11 +8,7 @@ import AgentLoop from '@deepseek-ai/dsh-agent-loop' import * as JsonlSessionPersistence from '@deepseek-ai/dsh-session-persistence-jsonl' import { mountAgentLoopTestDependencies } from '@deepseek-ai/dsh-agent-loop-testkit' import { SessionId } from '@deepseek-ai/dsh-session' -import { bindScopeParent, createScope, scopeOf } from '@deepseek-ai/dsh-scope' import LocalFileReference from '@deepseek-ai/dsh-file-reference-local' -import SubagentRuntime from '@deepseek-ai/dsh-subagent' -import * as SubagentSpawn from '@deepseek-ai/dsh-subagent-spawn-in-process' -import * as SubagentTool from '@deepseek-ai/dsh-tool-subagent' import Selection from '@deepseek-ai/dsh-tool-subagent/model-selection-settings' import type { Agent } from '@deepseek-ai/dsh-agent' import { createUserMessage } from '@deepseek-ai/dsh-llm' @@ -98,19 +94,12 @@ describe('serial creation listener integrations', () => { } }) - it('rolls back creation when the shared preset cannot capture model-selection policy', async () => { + it('rolls back creation when the Host settings service cannot capture model-selection policy', async () => { const ctx = await core() await ctx.plugin(Selection, { enabled: true, allowedModels: [] }) - await ctx.plugin(SubagentRuntime) - await ctx.plugin(SubagentSpawn, { providerName: 'spawn' }) - const preset = createScope(ctx, { preset: 'review' }) - await preset.ctx.plugin(SubagentTool, { provider: 'spawn', modelSelectionSettings: true }) try { await expect(ctx.agents.create({ sessionId: SessionId('review-tool'), - setup(agentCtx) { - bindScopeParent(scopeOf(agentCtx)!, scopeOf(preset.ctx)!) - }, })).rejects.toThrow('subagent model selection requires at least one allowed model') expect(ctx.agents.list()).toEqual([]) expect(ctx.sessions.list()).toEqual([]) diff --git a/packages/extensions/tool-cordis/src/api-catalog.ts b/packages/extensions/tool-cordis/src/api-catalog.ts index 4c2409e280..87e63e1432 100644 --- a/packages/extensions/tool-cordis/src/api-catalog.ts +++ b/packages/extensions/tool-cordis/src/api-catalog.ts @@ -2900,12 +2900,12 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [ }, { key: 'subagentModelSelection', - summary: 'Singleton settings owner read when delegation tools are composed for a Session.', - description: 'Singleton settings owner read when delegation tools are composed for a Session.', + summary: 'Host settings owner that records Session authorization before the first model request.', + description: 'Host settings owner that records Session authorization before the first model request.', methods: [ { signature: 'current(): SubagentModelSelectionSettings', - description: 'Read a detached selection preference for the next eligible Session composition.', + description: 'Read a detached selection preference for the next fresh top-level Session.', parameters: [], returns: 'the enabled state and exact allowed routes.', }, diff --git a/packages/subagent/subagent-dsh-sdk/tests/fixtures/loader/scoped-tool-subagent.ts b/packages/subagent/subagent-dsh-sdk/tests/fixtures/loader/scoped-tool-subagent.ts index cd30075474..71c9b7531c 100644 --- a/packages/subagent/subagent-dsh-sdk/tests/fixtures/loader/scoped-tool-subagent.ts +++ b/packages/subagent/subagent-dsh-sdk/tests/fixtures/loader/scoped-tool-subagent.ts @@ -3,6 +3,7 @@ import type { Context } from '@deepseek-ai/cordis' import type { Agent } from '@deepseek-ai/dsh-agent' import * as ToolSubagent from '@deepseek-ai/dsh-tool-subagent' +import * as ListModels from '@deepseek-ai/dsh-tool-subagent/list-models' import type { Config } from '@deepseek-ai/dsh-tool-subagent' export const name = 'scoped-tool-subagent' @@ -15,16 +16,17 @@ export const inject = ['agents', 'subagentModelSelection'] */ export function apply(ctx: Context, config: Config): void { const install = (agent: Agent): void => { + agent.ctx.plugin(ListModels) agent.ctx.inject(ToolSubagent.inject, (runtimeCtx) => { ToolSubagent.apply(runtimeCtx, { provider: config.provider, - modelSelectionSettings: true, + modelSelection: true, ...(config.toolName === undefined ? {} : { toolName: config.toolName }), ...(config.agentOptions === undefined ? {} : { agentOptions: config.agentOptions }), ...(config.persona === undefined ? {} : { persona: config.persona }), ...(config.toolFilter === undefined ? {} : { toolFilter: config.toolFilter }), ...(config.maxDepth === undefined ? {} : { maxDepth: config.maxDepth }), - }, agent.session) + }) }) } ctx.on('agent/created', ({ agent }) => { install(agent) }) diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index f02f435211..d5a9d02abb 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -18,20 +18,20 @@ en: 93f1701f5209ecb5 zh: 6310fd012400aead /deepseek-ai-dsh-tool-subagent/use-this-package/minimal-configuration: - en: 5465759f3c7fd5d7 - zh: 3f0b0cba46dbbc08 + en: f59fcf460bf140c9 + zh: e352bb50015b3a0f /deepseek-ai-dsh-tool-subagent/use-this-package/managed-delegation: en: 42673064ba88a211 zh: 8fa26285bd649dc9 /deepseek-ai-dsh-tool-subagent/use-this-package/selecting-a-child-llm: - en: ff7357f058f309ba - zh: 489f05233c1785a7 + en: a13c6fe5ac74be17 + zh: 3e06bb4e9662a263 /deepseek-ai-dsh-tool-subagent/understand-the-implementation: en: 6ce5f749256a7e6c zh: a4f2adb04a70d9cf /deepseek-ai-dsh-tool-subagent/understand-the-implementation/design-concept: - en: 794a71280d73dbc2 - zh: acb6990a599f3f66 + en: 5afcdbd58d405c29 + zh: ec11baf316a4e208 /deepseek-ai-dsh-tool-subagent/understand-the-implementation/activation-ownership: en: ef692c4d17290920 zh: b1d0d3dab011e2cd @@ -54,11 +54,11 @@ en: e4c7376d4f29dbd5 zh: 6b7a9eaf69c9ea53 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/what-the-model-sees: - en: 8266f5b60ac830db - zh: 6dceefd04897b6b0 + en: 126ffa0f945dd41a + zh: 26b5e8d700c3447c /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/token-effect: - en: 86d1dafcb6c00c57 - zh: fe890e9c6146c45e + en: 5df8d3c13b5a4c54 + zh: b2fb1eb115d38ac1 /deepseek-ai-dsh-tool-subagent/model-experience/tool-schema/kv-cache-effect: en: 5697c49cb84fb2a4 zh: b2b8464d3cbabfa5 @@ -66,11 +66,11 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: d0df5c0c9c988408 - zh: cadaa5591dcc5784 + en: 192670f899eb74f5 + zh: 06eac751268c3a9d /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: - en: cdb02d92cc19062e - zh: c466d4592828021f + en: 4fa61dc9a52259a2 + zh: bc51ac3b1a1cc6ad /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/kv-cache-effect: en: 7dc47a34d5574fbd zh: bd7bd2a4da2f8ae2 @@ -114,8 +114,8 @@ en: f5d5f97a153f7c70 zh: 3d6bc05b9eadfbc2 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work: - en: cbe53030678b9dd4 - zh: 06ccb6828e5708d5 + en: 637b5efa701c25f5 + zh: 5e00b602050780b6 /deepseek-ai-dsh-tool-subagent/known-limitations-and-deferred-work/dev-note: en: e67f633dc6fe7773 zh: 3bb0c20c93f16f98 diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index b8de1764f4..5e14fb9b42 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -46,7 +46,7 @@ Load the subagent service, an in-process or remote backend, and this tool; then |---|---|---| | `provider` | required | Provider name on `ctx.subagents` (e.g. `spawn`, `fork`, `acp`) | | `toolName` | `subagent` | Model-facing tool name; distinct for every loaded instance | -| `modelSelectionSettings` | `false` | Expose child route fields and `list_subagent_models`, and sample the Host's exact-route authorization preference for each top-level Session; a standing preset observes matching Sessions, while direct Agent setup passes its Session explicitly; requires provider `agentOptions` support | +| `modelSelection` | `false` | Expose optional child route fields; requires provider `agentOptions` support | | `agentOptions` | — | Configured child `provider`, `model`, adapter-owned `reasoningEffort`, and positive `maxTokens` defaults; requires provider `agentOptions` support and overlays any provider-owned route defaults | | `persona` | — | Per-child persona; requires the provider's `persona` capability | | `toolFilter` | — | Per-child global-tool restriction; requires the `toolFilter` capability | @@ -64,7 +64,7 @@ Every backend’s completion notice carries its final answer. Local Spawn and Fo ### Selecting a child LLM -Set `modelSelectionSettings: true` to expose optional `provider`, `model`, and `reasoning_effort` fields and register `list_subagent_models` with this instance; instances without it expose neither and reject selection fields. The instance samples the Host’s `subagent-model-selection` preference for fresh top-level Sessions. When enabled, its exact route list is recorded in the Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. This mode requires a backend that advertises `agentOptions`; in-process Spawn and Fork and DSH SDK qualify, while ACP, Codex, and Claude Code fail the mount. +Set `modelSelection: true` to expose optional `provider`, `model`, and `reasoning_effort` fields; instances without it reject selection fields. Mount `@deepseek-ai/dsh-tool-subagent/list-models` independently to expose `list_subagent_models`. Web and Desktop presets mount discovery by default, enable route fields for spawn, and leave them disabled for fork. The Host’s `/model-selection-settings` service samples the user preference during Agent creation, independently of tool installation. When enabled, its exact route list is recorded in each fresh top-level Session, inherited by child Sessions, and preserved on restore; later settings edits affect only new Sessions. Without a recorded Session policy, discovery and explicit selection permit only the parent Agent’s current effective provider/model and its supported efforts. Route fields require a backend that advertises `agentOptions`; in-process Spawn and Fork and DSH SDK qualify, while ACP, Codex, and Claude Code fail the mount. A call supplies `provider` and `model` together, or supplies only an effort when configured, parent, or provider-owned defaults provide the route. Static `provider.agentRouteDefaults`, when present, form the provider/model baseline; tool configuration and model fields overlay it before route-aware effort merging and exact-route preflight. Providers without these defaults use compatible values from the parent's latest logged request, then the parent's creation options before its first request, while retaining the configured `maxTokens`. Changing the route without an explicit effort clears the inherited route-owned effort, so the selected model resolves its default. The live LLM adapter validates the effective route before child creation. Catalog membership remains advisory, so a model can use an unlisted id when its adapter accepts it. @@ -80,7 +80,7 @@ This section explains how the tool mirrors provider lifecycle and settles runs; ### Design concept -One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Settings-backed instances also register `list_subagent_models` in the same composition and capture Session authorization during Agent creation; calls read the durable policy independently of registration. Direct Agent setup supplies its unpublished Session explicitly. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. At most one instance in a tool scope may set `modelSelectionSettings` because `list_subagent_models` has a global name. No runtime invariant companion is published: policy comes directly from the Session projection and is checked at execution. +One instance is one provider plus one tool name. The plugin registers its definition in the owning composition while the provider is present and disposes it when the provider leaves. Agents inherit that shared definition, so their existing tool filters apply without per-Agent registration. Discovery is a separate plugin and reads the same durable Session policy as the delegation executor. A numeric `maxDepth` or configured LLM selection unsupported by the provider fails the mount. Multiple distinctly named delegation tools may expose model selection in one scope. ### Activation ownership @@ -128,11 +128,11 @@ Read these pages when the package-level contract is not enough; they move from t #### What the model sees -The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. An instance with `modelSelectionSettings` adds `provider`, `model`, and `reasoning_effort` plus route guidance in every Session. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. +The generated default [`subagent` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent) appears under the configured tool name while its provider exists. An instance with `modelSelection` adds `provider`, `model`, and `reasoning_effort` plus route guidance in every Session. Context inheritance and continuation support determine task and follow-up descriptions. A tool restriction removes both the schema and its `tool:` guidance section. #### Token effect -Fixed schema cost per parent request; `modelSelectionSettings` adds three optional route fields. Each tool instance adds one schema; visible delegation tools share one short system-prompt section. +Fixed schema cost per parent request; `modelSelection` adds three optional route fields. Each tool instance adds one schema; visible delegation tools share one short system-prompt section. #### KV Cache effect @@ -142,11 +142,11 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -The `list_subagent_models` definition registered by a settings-backed instance reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. +The independent `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. #### Token effect -One fixed discovery schema is present in compositions with `modelSelectionSettings`, regardless of the Session’s preference; the shipped Web presets include it, while headless and SDK profiles do not. Route metadata enters the transcript only when the model calls the tool. +One fixed discovery schema is present wherever `/list-models` is mounted, regardless of the Session’s preference or delegation tools; the shipped Web and Desktop presets include it, while headless and SDK profiles do not. Route metadata enters the transcript only when the model calls the tool. #### KV Cache effect @@ -208,8 +208,8 @@ Append-only; newly visible content follows the reusable request prefix and does These limits define what this tool does not return or enforce; they are current package constraints. - **This tool does not wait for results** — use the completion notice or child messages. Programmatic result waiting belongs to the subagent service. -- **Shipped fork tools omit model selection** — their configuration leaves `modelSelectionSettings` disabled to preserve inherited-prefix reuse. Custom fork tools may enable it under the same Session authorization and adapter validation as other delegation tools. -- **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection requires `modelSelectionSettings` and is restricted to the Session’s authorized routes, or the current parent route without an enabled policy. +- **Shipped fork tools omit model selection** — their configuration leaves `modelSelection` disabled to preserve inherited-prefix reuse. Custom fork tools may enable it under the same Session authorization and adapter validation as other delegation tools. +- **Non-routing child policy is fixed per instance** — another persona, tool filter, or depth cap requires another distinctly named tool. Explicit LLM selection requires `modelSelection` and is restricted to the Session’s authorized routes, or the current parent route without an enabled policy. ### Dev Note diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index 575b9b4530..e8d19608f2 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -46,7 +46,7 @@ kind: "package-reference" |---|---|---| | `provider` | 必填 | `ctx.subagents` 上的提供方名称（如 `spawn`、`fork`、`acp`） | | `toolName` | `subagent` | 面向模型的工具名称；每个已加载实例必须不同 | -| `modelSelectionSettings` | `false` | 公开子级路由字段和 `list_subagent_models`，并为每个顶层 Session 读取宿主的精确路由授权偏好；常驻 preset 观察匹配 Session，直接 Agent setup 则显式传入其 Session；要求提供方支持 `agentOptions` | +| `modelSelection` | `false` | 公开可选子级路由字段；要求提供方支持 `agentOptions` | | `agentOptions` | — | 配置的子级 `provider`、`model`、适配器所有的 `reasoningEffort` 与正整数 `maxTokens` 默认值；要求提供方支持 `agentOptions`，并会覆盖提供方持有的路由默认值 | | `persona` | — | 每个子 agent 独立的 persona；要求提供方具备 `persona` 能力 | | `toolFilter` | — | 每个子 agent 独立的全局工具限制；要求提供方具备 `toolFilter` 能力 | @@ -64,7 +64,7 @@ kind: "package-reference" ### 选择子级 LLM -设置 `modelSelectionSettings: true` 后，该实例公开可选的 `provider`、`model` 和 `reasoning_effort` 字段，并注册 `list_subagent_models`；未设置的实例两者都不公开，并拒绝选择字段。该实例为新顶层 Session 读取 Host 的 `subagent-model-selection` 偏好。启用时，精确路由列表记录在 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。此模式要求后端公布 `agentOptions`；进程内 Spawn、Fork 和 DSH SDK 满足条件，ACP、Codex 和 Claude Code 会使挂载失败。 +设置 `modelSelection: true` 后，该实例公开可选的 `provider`、`model` 和 `reasoning_effort` 字段；未设置的实例拒绝选择字段。独立挂载 `@deepseek-ai/dsh-tool-subagent/list-models` 以公开 `list_subagent_models`。Web 与 Desktop preset 默认挂载发现工具，为 spawn 启用路由字段，为 fork 保持关闭。Host 的 `/model-selection-settings` 服务在 Agent 创建时读取用户偏好，与工具安装无关。启用时，精确路由列表记录在每个新顶层 Session 中，由子 Session 继承，并在恢复时保留；之后的设置编辑只影响新 Session。没有已记录 Session 策略时，发现和显式选择只允许父 Agent 当前生效的 provider/model 及其支持的强度。路由字段要求后端公布 `agentOptions`；进程内 Spawn、Fork 和 DSH SDK 满足条件，ACP、Codex 和 Claude Code 会使挂载失败。 一次调用需同时提供 `provider` 与 `model`；当配置值、父 agent 值或提供方持有的默认值能提供路由时，也可只提供推理等级。静态的 `provider.agentRouteDefaults` 在存在时构成提供方／模型基线；工具配置与模型字段会在路由相关强度合并和确切路由预检前覆盖它。没有这些默认值的提供方会使用父 agent 最新已记录请求中的兼容值，再使用父级首次请求前的创建选项，并保留配置的 `maxTokens`。更改路由但未显式提供推理等级时，会清除继承的路由自有等级，使所选模型解析自己的默认值。实时 LLM 适配器在创建子 agent 前校验有效路由。目录成员资格只提供建议，因此适配器接受时，模型可以使用未列出的 id。 @@ -80,7 +80,7 @@ kind: "package-reference" ### 设计理念 -一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。读取设置的实例还会在同一组合中注册 `list_subagent_models`，并在 Agent 创建时记录 Session 授权；调用时独立于注册读取持久策略。直接配置 Agent 时显式传入尚未发布的 Session。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。由于 `list_subagent_models` 使用全局名称，一个工具作用域中最多只有一个实例可以设置 `modelSelectionSettings`。本包不导出运行时不变量：策略直接来自 Session 投影，并在执行时校验。 +一个实例对应一个提供方和一个工具名。插件在提供方存在时向所属组合注册定义，并在提供方离开时释放它。Agent 继承这份共享定义，因此现有工具过滤器即可生效，无需逐 Agent 注册。发现工具是独立插件，与委派执行器读取同一份持久 Session 策略。提供方不支持的数字 `maxDepth` 或配置的 LLM 选择会使挂载失败。同一作用域中可以有多个名称不同的委派工具公开模型选择。 ### Activation 所有权 @@ -128,11 +128,11 @@ kind: "package-reference" #### 模型看到什么 -生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。设置了 `modelSelectionSettings` 的实例在每个 Session 中都会增加 `provider`、`model`、`reasoning_effort` 及路由指导。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 +生成的默认 [`subagent` schema](../../../docs/tool-catalog.zh.md#deepseek-aidsh-tool-subagent) 在提供方存在时以配置的工具名公开。设置了 `modelSelection` 的实例在每个 Session 中都会增加 `provider`、`model`、`reasoning_effort` 及路由指导。上下文继承和继续执行能力决定任务与后续工作描述。工具限制同时移除 schema 及其 `tool:` 指导段。 #### Token 影响 -每次父级请求承担固定 schema 成本；`modelSelectionSettings` 会增加三个可选路由字段。每个工具实例增加一份 schema；可见委派工具共享一个简短系统提示词段。 +每次父级请求承担固定 schema 成本；`modelSelection` 会增加三个可选路由字段。每个工具实例增加一份 schema；可见委派工具共享一个简短系统提示词段。 #### KV Cache 影响 @@ -142,11 +142,11 @@ kind: "package-reference" #### 模型看到什么 -读取设置的实例注册的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 +独立的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 #### Token 影响 -设置了 `modelSelectionSettings` 的组合中存在一份固定发现 schema，不受 Session 偏好影响；随附的 Web preset 包含它，headless 与 SDK profile 不包含。只有模型调用工具时，路由元数据才进入 transcript。 +挂载 `/list-models` 的组合中存在一份固定发现 schema，不受 Session 偏好或委派工具影响；随附 Web 与 Desktop preset 包含它，headless 与 SDK profile 不包含。只有模型调用工具时，路由元数据才进入 transcript。 #### KV Cache 影响 @@ -208,8 +208,8 @@ Start independent delegations with `subagent` or `subagent_fork` together in one 这些限制说明本工具不返回或不强制执行什么；它们是当前包约束。 - **本工具不等待结果**——使用完成通知或子任务消息。程序化结果等待由 subagent 服务提供。 -- **随附 fork 工具省略模型选择**：其配置保持 `modelSelectionSettings` 关闭，以保留对继承前缀的复用。自定义 fork 工具可以启用它，并接受与其他委派工具相同的 Session 授权和适配器校验。 -- **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择需要 `modelSelectionSettings`，并受 Session 授权路由限制，没有启用策略时只允许当前父级路由。 +- **随附 fork 工具省略模型选择**：其配置保持 `modelSelection` 关闭，以保留对继承前缀的复用。自定义 fork 工具可以启用它，并接受与其他委派工具相同的 Session 授权和适配器校验。 +- **非路由子级策略按实例固定** — 更换角色、工具过滤器或深度上限，需要另一个不同名称的工具。显式 LLM 选择需要 `modelSelection`，并受 Session 授权路由限制，没有启用策略时只允许当前父级路由。 ### 开发备注 diff --git a/packages/subagent/tool-subagent/package.json b/packages/subagent/tool-subagent/package.json index a535e39446..00f6896dc0 100644 --- a/packages/subagent/tool-subagent/package.json +++ b/packages/subagent/tool-subagent/package.json @@ -25,13 +25,18 @@ "./src/*": "./src/*", "./package.json": "./package.json", "./locale/*.json": "./locale/*.json", - "./model-selection-settings/locale/*.json": "./locale/model-selection-settings/*.json" + "./model-selection-settings/locale/*.json": "./locale/model-selection-settings/*.json", + "./list-models": { + "types": "./lib/types/list-models.d.ts", + "default": "./lib/list-models.js" + } }, "files": [ "locale/*.json", "locale/model-selection-settings/*.json", "lib/index.js", "lib/model-selection-settings.js", + "lib/list-models.js", "lib/types/**/*.d.ts" ], "license": "MIT", diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index d07e4aa883..e6185ce580 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -8,14 +8,11 @@ import type { Context } from '@deepseek-ai/cordis' import z from '@deepseek-ai/schemastery' -import { scopeChainOf, scopeOf } from '@deepseek-ai/dsh-scope' import { defineTool } from '@deepseek-ai/dsh-tools' import type { ToolDefinition } from '@deepseek-ai/dsh-tools' import type { AgentOptions } from '@deepseek-ai/dsh-agent' import { ReasoningEffortId } from '@deepseek-ai/dsh-llm' import type { ContentBlock } from '@deepseek-ai/dsh-llm' -import { SessionSeq } from '@deepseek-ai/dsh-session' -import type { Session } from '@deepseek-ai/dsh-session' import { assertSubagentMaxDepth, parentAgentOptionsForDelegation, @@ -29,10 +26,7 @@ import { requestedAgentOptions, } from './model-selection.ts' import type { DelegationModelRequest } from './model-selection.ts' -import { registerListSubagentModels } from './list-models.ts' -import type {} from './model-selection-settings.ts' import { - recordSubagentModelSelection, subagentModelSelectionProjectionDefinition, subagentModelSelectionPolicy, } from './model-selection-state.ts' @@ -54,13 +48,11 @@ export interface Config { */ toolName?: string /** - * Expose optional child `provider`, `model`, and `reasoning_effort` fields - * plus `list_subagent_models`, and sample the Host `subagent-model-selection` - * setting for each new top-level Session; child Sessions inherit that - * decision. Requires a provider with `agentOptions` support. - * At most one instance per tool scope may enable it. + * Expose optional child `provider`, `model`, and `reasoning_effort` fields. + * Requires provider `agentOptions` support. Execution enforces the Session + * policy, or restricts explicit selection to the current parent route. */ - modelSelectionSettings?: boolean + modelSelection?: boolean /** * Agent options applied to every child; omitted fields use child-loop defaults. */ @@ -97,7 +89,7 @@ export interface Config { export const Config: z = z.object({ provider: z.string().required(), toolName: z.string().default('subagent'), - modelSelectionSettings: z.boolean().default(false), + modelSelection: z.boolean().default(false), // Prevent Schemastery from materializing omitted agentOptions as `{}`. agentOptions: z.object({ provider: z.string(), @@ -161,9 +153,8 @@ function providerWording(inheritsConversation: boolean): { description: string; * Install one delegation-tool composition. * @param ctx - Context that owns the registrations. * @param config - delegation-tool configuration. - * @param session - unpublished Session supplied by a direct Agent setup; omit for a standing composition. */ -export function apply(ctx: Context, config: Config, session?: Session): void { +export function apply(ctx: Context, config: Config): void { // Direct apply() bypasses Schemastery's numeric constraints. A direct-apply // omission stays capless (the schema default only runs through the loader). if (config.maxDepth !== 'provider-managed') assertSubagentMaxDepth(config.maxDepth) @@ -173,7 +164,7 @@ export function apply(ctx: Context, config: Config, session?: Session): void { } const toolName = config.toolName ?? 'subagent' - const modelSelectionEnabled = config.modelSelectionSettings === true + const modelSelectionEnabled = config.modelSelection === true ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) const assertSubagentProviderConfiguration = (subagentProvider: SubagentProvider): void => { @@ -203,253 +194,173 @@ export function apply(ctx: Context, config: Config, session?: Session): void { const initialProvider = ctx.subagents.getProvider(config.provider) if (initialProvider !== undefined) assertSubagentProviderConfiguration(initialProvider) - const install = (runtimeCtx: Context): void => { - if (modelSelectionEnabled) { - registerListSubagentModels(runtimeCtx, (exec) => { - if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') - const routes = subagentModelSelectionPolicy(runtimeCtx.sessionProjections, exec.agent.session) - return routes === undefined ? undefined : { routes } - }) - } - // Load order and HMR replacement can change provider availability while - // this fiber remains active. - let mounted: { subagentProvider: SubagentProvider; disposeTool: () => void } | undefined - const mount = (subagentProvider: SubagentProvider): void => { - assertSubagentProviderConfiguration(subagentProvider) - const wording = providerWording(subagentProvider.inheritsParentContext) - const providerRouteDefaults = subagentProvider.agentRouteDefaults - const selectionDescription = providerRouteDefaults !== undefined - ? ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend\'s route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' - : ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' - const definition = defineTool({ - name: toolName, - description: wording.description - + ' This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes.' - + (subagentProvider.prepareContinuable !== undefined - ? ' The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.' - : ' The completion notice includes its final answer. This backend does not accept follow-up messages.') - + (modelSelectionEnabled ? selectionDescription : ''), - parameters: { - cwd: { - type: 'string', - description: 'Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent.', + // Load order and HMR replacement can change provider availability while + // this fiber remains active. + let mounted: { subagentProvider: SubagentProvider; disposeTool: () => void } | undefined + const mount = (subagentProvider: SubagentProvider): void => { + assertSubagentProviderConfiguration(subagentProvider) + const wording = providerWording(subagentProvider.inheritsParentContext) + const providerRouteDefaults = subagentProvider.agentRouteDefaults + const selectionDescription = providerRouteDefaults !== undefined + ? ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend\'s route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' + : ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' + const definition = defineTool({ + name: toolName, + description: wording.description + + ' This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes.' + + (subagentProvider.prepareContinuable !== undefined + ? ' The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.' + : ' The completion notice includes its final answer. This backend does not accept follow-up messages.') + + (modelSelectionEnabled ? selectionDescription : ''), + parameters: { + cwd: { + type: 'string', + description: 'Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent.', + }, + description: { + type: 'string', + required: true, + description: 'A short (3-5 word) description of the delegated task, for display.', + }, + prompt: { + type: 'string', + required: true, + description: wording.promptDescription, + }, + ...modelSelectionEnabled ? { + provider: { + type: 'string' as const, + description: 'Child LLM provider id.', }, - description: { - type: 'string', - required: true, - description: 'A short (3-5 word) description of the delegated task, for display.', + model: { + type: 'string' as const, + description: 'Model id within that provider.', }, - prompt: { - type: 'string', - required: true, - description: wording.promptDescription, + reasoning_effort: { + type: 'string' as const, + description: 'Model-specific reasoning effort id.', }, - ...modelSelectionEnabled ? { - provider: { - type: 'string' as const, - description: 'Child LLM provider id.', - }, - model: { - type: 'string' as const, - description: 'Model id within that provider.', - }, - reasoning_effort: { - type: 'string' as const, - description: 'Model-specific reasoning effort id.', - }, - } : {}, + } : {}, - }, - output: { - schema: { - type: 'object', - additionalProperties: false, - properties: { - kind: { type: 'string', required: true, const: 'activation' }, - subagentId: { type: 'string', required: true }, - }, + }, + output: { + schema: { + type: 'object', + additionalProperties: false, + properties: { + kind: { type: 'string', required: true, const: 'activation' }, + subagentId: { type: 'string', required: true }, }, - render: (_args, value) => [{ type: 'text', text: `started subagent ${value.subagentId}` }], }, - isConcurrencySafe: () => true, - async execute(args, exec) { - const parent = exec.agent - if (!parent) { - // Non-agent callers provide no parent for delegation ownership. - throw new Error('subagent tool requires a calling agent (exec.agent was undefined)') + render: (_args, value) => [{ type: 'text', text: `started subagent ${value.subagentId}` }], + }, + isConcurrencySafe: () => true, + async execute(args, exec) { + const parent = exec.agent + if (!parent) { + // Non-agent callers provide no parent for delegation ownership. + throw new Error('subagent tool requires a calling agent (exec.agent was undefined)') + } + + const routes = subagentModelSelectionPolicy(ctx.sessionProjections, parent.session) + const modelSelectionPolicy = routes === undefined ? undefined : { routes } + const modelRequest = args as DelegationModelRequest + const parentOptions = parentAgentOptionsForDelegation(parent) + const requiresRoutePreflight = hasDelegationModelRequest(modelRequest) + || hasConfiguredLlmSelection(config.agentOptions) + const configuredChildAgentOptions = requiresRoutePreflight && providerRouteDefaults !== undefined + ? { ...providerRouteDefaults, ...config.agentOptions } + : config.agentOptions + const requestedChildAgentOptions = requestedAgentOptions( + parentOptions, + configuredChildAgentOptions, + modelRequest, + modelSelectionEnabled, + ) + assertAllowedModelSelection( + modelSelectionPolicy, + parentOptions, + requestedChildAgentOptions, + modelRequest, + ) + if (requiresRoutePreflight) { + const llm = ctx.get('llm') + if (llm === undefined) { + throw new Error('cannot resolve the selected child LLM route because the `llm` service is unavailable') } - - const routes = subagentModelSelectionPolicy(runtimeCtx.sessionProjections, parent.session) - const modelSelectionPolicy = routes === undefined ? undefined : { routes } - const modelRequest = args as DelegationModelRequest - const parentOptions = parentAgentOptionsForDelegation(parent) - const requiresRoutePreflight = hasDelegationModelRequest(modelRequest) - || hasConfiguredLlmSelection(config.agentOptions) - const configuredChildAgentOptions = requiresRoutePreflight && providerRouteDefaults !== undefined - ? { ...providerRouteDefaults, ...config.agentOptions } - : config.agentOptions - const requestedChildAgentOptions = requestedAgentOptions( - parentOptions, - configuredChildAgentOptions, - modelRequest, - modelSelectionEnabled, - ) - assertAllowedModelSelection( - modelSelectionPolicy, + await preflightChildLlmRoute( + llm, parentOptions, requestedChildAgentOptions, - modelRequest, + exec.signal, + providerRouteDefaults === undefined, ) - if (requiresRoutePreflight) { - const llm = runtimeCtx.get('llm') - if (llm === undefined) { - throw new Error('cannot resolve the selected child LLM route because the `llm` service is unavailable') - } - await preflightChildLlmRoute( - llm, - parentOptions, - requestedChildAgentOptions, - exec.signal, - providerRouteDefaults === undefined, - ) - if (runtimeCtx.subagents.getProvider(config.provider) !== subagentProvider) { - throw new Error(`subagent provider "${config.provider}" changed while resolving the child LLM route; retry the delegation`) - } + if (ctx.subagents.getProvider(config.provider) !== subagentProvider) { + throw new Error(`subagent provider "${config.provider}" changed while resolving the child LLM route; retry the delegation`) } - exec.signal.throwIfAborted() - const maxDepth = runtimeCtx.subagents.resolveMaxDepth(config.maxDepth) - const request = { - ...args.cwd === undefined ? {} : { cwd: args.cwd }, - prompt: [{ type: 'text', text: args.prompt }] as ContentBlock[], - parent, - ...requestedChildAgentOptions !== undefined ? { agentOptions: requestedChildAgentOptions } : {}, - ...config.persona !== undefined ? { persona: config.persona } : {}, - ...config.toolFilter !== undefined ? { toolFilter: config.toolFilter } : {}, - ...maxDepth !== undefined ? { maxDepth } : {}, - } - - const started = await runtimeCtx.subagents.startActivation({ - provider: config.provider, - label: args.description, - request, - signal: exec.signal, - delivery: 'parent', - }) - return { kind: 'activation' as const, subagentId: started.childId } - }, - }) - const disposeTool = runtimeCtx.effect(() => { - const unregister = runtimeCtx.tools.register(definition) - delegationTools.add(definition) - return () => { - delegationTools.delete(definition) - unregister() } - }) - // oxlint-disable-next-line typescript/no-misused-promises -- Tool and guidance disposal are synchronous. - mounted = { subagentProvider, disposeTool } - } + exec.signal.throwIfAborted() + const maxDepth = ctx.subagents.resolveMaxDepth(config.maxDepth) + const request = { + ...args.cwd === undefined ? {} : { cwd: args.cwd }, + prompt: [{ type: 'text', text: args.prompt }] as ContentBlock[], + parent, + ...requestedChildAgentOptions !== undefined ? { agentOptions: requestedChildAgentOptions } : {}, + ...config.persona !== undefined ? { persona: config.persona } : {}, + ...config.toolFilter !== undefined ? { toolFilter: config.toolFilter } : {}, + ...maxDepth !== undefined ? { maxDepth } : {}, + } - // Register listeners before checking presence so no synchronous change is missed. - runtimeCtx.on('subagent/provider-added', (subagentProvider) => { - if (subagentProvider.name === config.provider && mounted === undefined) mount(subagentProvider) - }) - runtimeCtx.on('subagent/provider-removed', (name) => { - if (name !== config.provider || mounted === undefined) return - mounted.disposeTool() - mounted = undefined - }) - const present = runtimeCtx.subagents.getProvider(config.provider) - if (present !== undefined) { - mount(present) - } else { - // A backend fiber may activate later; a misspelled provider remains visible in this log. - runtimeCtx.logger.info(`subagent provider "${config.provider}" not registered yet; the "${config.toolName ?? 'subagent'}" tool will register when it appears`) - } - runtimeCtx.systemPrompt.section({ - name: `tool:${toolName}`, - order: runtimeCtx.systemPrompt.getSectionOrder('TOOL_SUBAGENT'), - text: (context) => { - if (mounted === undefined) return '' - const visible = [...delegationTools] - .filter(tool => runtimeCtx.tools.get(tool.name, context.scope) === tool) - .map(tool => tool.name) - .sort() - if (visible[0] !== toolName) return '' - const names = visible.map(name => `\`${name}\``).join(' or ') - return `Start independent delegations with ${names} together in one assistant message and continue useful work while they run.` + const started = await ctx.subagents.startActivation({ + provider: config.provider, + label: args.description, + request, + signal: exec.signal, + delivery: 'parent', + }) + return { kind: 'activation' as const, subagentId: started.childId } }, }) - } - - if (config.modelSelectionSettings !== true) { - install(ctx) - return - } - - const settings = ctx.get('subagentModelSelection') - if (settings === undefined) { - throw new Error( - 'tool-subagent: `modelSelectionSettings` requires ' - + '@deepseek-ai/dsh-tool-subagent/model-selection-settings in the Host scope', - ) - } - const selectForSession = (target: Session): void => { - const freshSession = target.firstLiveSeq === 0 - // oxlint-disable-next-line typescript/no-deprecated -- Existing Session history read; migration deferred. - && target.eventAt(SessionSeq(0))?.type !== 'session/end-seed' - let allowedModels = subagentModelSelectionPolicy(ctx.sessionProjections, target) - if (allowedModels === undefined) { - const parentId = target.header.origin === 'subagent' - ? target.header.parentSession - : undefined - if (parentId !== undefined) { - const sessions = ctx.get('sessions') - if (sessions === undefined) { - throw new Error('tool-subagent: child model-selection inheritance requires the Session registry') - } - const parent = sessions.get(parentId) - allowedModels = parent === undefined - ? undefined - : subagentModelSelectionPolicy(ctx.sessionProjections, parent) - } else if (freshSession) { - const current = settings.current() - allowedModels = current.enabled ? current.allowedModels : undefined + const disposeTool = ctx.effect(() => { + const unregister = ctx.tools.register(definition) + delegationTools.add(definition) + return () => { + delegationTools.delete(definition) + unregister() } - } - if (allowedModels !== undefined) { - recordSubagentModelSelection(ctx.sessionProjections, target, allowedModels) - } + }) + // oxlint-disable-next-line typescript/no-misused-promises -- Tool and guidance disposal are synchronous. + mounted = { subagentProvider, disposeTool } } - if (session !== undefined) { - selectForSession(session) - install(ctx) - return + // Register listeners before checking presence so no synchronous change is missed. + ctx.on('subagent/provider-added', (subagentProvider) => { + if (subagentProvider.name === config.provider && mounted === undefined) mount(subagentProvider) + }) + ctx.on('subagent/provider-removed', (name) => { + if (name !== config.provider || mounted === undefined) return + mounted.disposeTool() + mounted = undefined + }) + const present = ctx.subagents.getProvider(config.provider) + if (present !== undefined) { + mount(present) + } else { + // A backend fiber may activate later; a misspelled provider remains visible in this log. + ctx.logger.info(`subagent provider "${config.provider}" not registered yet; the "${config.toolName ?? 'subagent'}" tool will register when it appears`) } - - const compositionScope = scopeOf(ctx) - if (compositionScope === undefined) { - throw new Error('tool-subagent: standing `modelSelectionSettings` requires a scoped preset Context') - } - const agents = ctx.get('agents') - /* v8 ignore next -- shipped preset compositions always include the Agent registry. */ - if (agents === undefined) throw new Error('tool-subagent: standing `modelSelectionSettings` requires the Agent registry') - const captured = new WeakSet() - const capture = (target: Session): void => { - if (captured.has(target)) return - selectForSession(target) - captured.add(target) - } - const captureComposedSessions = (): void => { - for (const agent of agents.list()) { - if (scopeChainOf(scopeOf(agent.ctx)).includes(compositionScope)) capture(agent.session) - } - } - ctx.on('agent/created', ({ agent }) => { capture(agent.session) }) - // Blank Sessions can join a preset after creation; absence also records the - // sampled off decision for this plugin lifetime. - ctx.on('tools/change', captureComposedSessions) - captureComposedSessions() - install(ctx) + ctx.systemPrompt.section({ + name: `tool:${toolName}`, + order: ctx.systemPrompt.getSectionOrder('TOOL_SUBAGENT'), + text: (context) => { + if (mounted === undefined) return '' + const visible = [...delegationTools] + .filter(tool => ctx.tools.get(tool.name, context.scope) === tool) + .map(tool => tool.name) + .sort() + if (visible[0] !== toolName) return '' + const names = visible.map(name => `\`${name}\``).join(' or ') + return `Start independent delegations with ${names} together in one assistant message and continue useful work while they run.` + }, + }) } diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index a25c9db8c8..4a81c568f6 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -7,6 +7,10 @@ import { defineTool } from '@deepseek-ai/dsh-tools' import { parentAgentOptionsForDelegation } from '@deepseek-ai/dsh-subagent' import type { ToolRunContext } from '@deepseek-ai/dsh-tools' import type { ModelSelectionPolicy } from './model-selection.ts' +import { subagentModelSelectionPolicy, subagentModelSelectionProjectionDefinition } from './model-selection-state.ts' + +export const name = 'tool-subagent-list-models' +export const inject = ['tools', 'sessionProjections'] interface ListSubagentModelsRequest { readonly provider?: string @@ -37,13 +41,14 @@ function modelLine(provider: string, model: { id: string; name: string; descript /** Read the requested provider, allowed models, or exact-model efforts. */ async function listSubagentModels( ctx: Context, - policy: ModelSelectionPolicy | undefined, request: ListSubagentModelsRequest, exec: ToolRunContext, ): Promise { + if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') + const routes = subagentModelSelectionPolicy(ctx.sessionProjections, exec.agent.session) + let policy = routes === undefined ? undefined : { routes } const selectionEnabled = policy !== undefined if (policy === undefined) { - if (exec.agent === undefined) throw new Error('list_subagent_models requires a calling agent') const { provider, model } = parentAgentOptionsForDelegation(exec.agent) if (provider === undefined || model === undefined) { throw new Error('cannot discover child LLM routes without a parent provider and model') @@ -94,12 +99,11 @@ async function listSubagentModels( } /** - * Register `list_subagent_models` for the delegation-tool instance that owns model selection. - * Provider listings intersect the catalog with authorization, or resolve only the parent model without a Session policy. - * @param ctx - Context whose tool registry owns the fixed discovery definition. - * @param policyFor - Read the calling Session’s authorized routes; undefined restricts discovery to its parent model. + * Register independent discovery of the calling Session’s authorized child models. + * @param ctx - Context carrying the tool registry and Session projections. */ -export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolRunContext) => ModelSelectionPolicy | undefined): void { +export function apply(ctx: Context): void { + ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) ctx.tools.register(defineTool({ name: 'list_subagent_models', description: @@ -121,7 +125,7 @@ export function registerListSubagentModels(ctx: Context, policyFor: (exec: ToolR render: (_args, result) => [{ type: 'text', text: result }], }, execute(args, exec) { - return listSubagentModels(ctx, policyFor(exec), args, exec) + return listSubagentModels(ctx, args, exec) }, })) } diff --git a/packages/subagent/tool-subagent/src/model-selection-settings.ts b/packages/subagent/tool-subagent/src/model-selection-settings.ts index e03d5d2b49..9e4907556b 100644 --- a/packages/subagent/tool-subagent/src/model-selection-settings.ts +++ b/packages/subagent/tool-subagent/src/model-selection-settings.ts @@ -3,6 +3,12 @@ import type { Volatile } from '@deepseek-ai/cordis' import { Context, Service } from '@deepseek-ai/cordis' import z from '@deepseek-ai/schemastery' +import { SessionSeq } from '@deepseek-ai/dsh-session' +import type {} from '@deepseek-ai/dsh-agent' +import { + subagentModelSelectionPolicy, + subagentModelSelectionProjectionDefinition, +} from './model-selection-state.ts' import { AllowedModelRouteSchema, assertAllowedModelRoutes, @@ -11,16 +17,16 @@ import { declare module '@deepseek-ai/cordis' { interface Context { - /** User preference sampled when a new Session receives delegation tools. */ + /** User preference sampled during Agent creation. */ subagentModelSelection: SubagentModelSelectionConfig } } /** Stored user preference; the shipped composition defaults it off. */ export interface SubagentModelSelectionSettings { - /** Whether newly composed top-level Sessions receive model selection. */ + /** Whether new top-level Sessions receive model selection. */ enabled: boolean - /** Exact child LLM routes offered to newly composed top-level Sessions. */ + /** Exact child LLM routes offered to new top-level Sessions. */ allowedModels: AllowedModelRoute[] } @@ -32,8 +38,10 @@ export interface Config { allowedModels: Volatile } -/** Singleton settings owner read when delegation tools are composed for a Session. */ +/** Host settings owner that records Session authorization before the first model request. */ export class SubagentModelSelectionConfig extends Service { + static inject = ['sessionProjections', 'sessions'] + static Config = z.object({ enabled: z.boolean().default(false).volatile(), allowedModels: z.array(AllowedModelRouteSchema).default([]).volatile(), @@ -41,10 +49,27 @@ export class SubagentModelSelectionConfig extends Service { constructor(ctx: Context, private config: Config) { super(ctx, 'subagentModelSelection') + ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) + ctx.on('agent/created', ({ agent }) => { + const target = agent.session + if (subagentModelSelectionPolicy(ctx.sessionProjections, target) !== undefined) return + let allowedModels: AllowedModelRoute[] | undefined + const parentId = target.header.origin === 'subagent' ? target.header.parentSession : undefined + if (parentId !== undefined) { + const parent = ctx.sessions.get(parentId) + if (parent !== undefined) allowedModels = subagentModelSelectionPolicy(ctx.sessionProjections, parent) + } else if (target.firstLiveSeq === 0 + // oxlint-disable-next-line typescript/no-deprecated -- Existing Session history read; migration deferred. + && target.eventAt(SessionSeq(0))?.type !== 'session/end-seed') { + const current = this.current() + allowedModels = current.enabled ? current.allowedModels : undefined + } + if (allowedModels !== undefined) target.append('subagent/model-selection-policy', { allowedModels }) + }) } /** - * Read a detached selection preference for the next eligible Session composition. + * Read a detached selection preference for the next fresh top-level Session. * @returns the enabled state and exact allowed routes. */ current(): SubagentModelSelectionSettings { diff --git a/packages/subagent/tool-subagent/src/model-selection-state.ts b/packages/subagent/tool-subagent/src/model-selection-state.ts index 0854425c83..221ef5a8f6 100644 --- a/packages/subagent/tool-subagent/src/model-selection-state.ts +++ b/packages/subagent/tool-subagent/src/model-selection-state.ts @@ -62,20 +62,3 @@ export function subagentModelSelectionPolicy( ): AllowedModelRoute[] | undefined { return projections.stateOf(session, 'subagentModelSelectionPolicy')?.map(route => ({ ...route })) } - -/** - * Append the route policy once, before the Session’s first model request. - * @param projections - registry that owns the policy projection. - * @param session - session receiving route authorization. - * @param allowedModels - exact routes the Session may select explicitly. - */ -export function recordSubagentModelSelection( - projections: Pick, - session: Session, - allowedModels: readonly AllowedModelRoute[], -): void { - if (subagentModelSelectionPolicy(projections, session) !== undefined) return - session.append('subagent/model-selection-policy', { - allowedModels: allowedModels.map(route => ({ ...route })), - }) -} diff --git a/packages/subagent/tool-subagent/tests/harness.ts b/packages/subagent/tool-subagent/tests/harness.ts index 01279f3b13..e8ae24cf4a 100644 --- a/packages/subagent/tool-subagent/tests/harness.ts +++ b/packages/subagent/tool-subagent/tests/harness.ts @@ -9,6 +9,7 @@ import SubagentRuntime from '@deepseek-ai/dsh-subagent' import { SessionId } from '@deepseek-ai/dsh-session' import * as mock from './scripted-provider.ts' import * as tool from '../src/index.ts' +import * as ListModels from '../src/list-models.ts' import SubagentModelSelectionConfig from '../src/model-selection-settings.ts' const contexts = new Set() @@ -51,12 +52,12 @@ const TEST_ALLOWED_MODELS = [ export async function setup(toolConfig: SetupConfig, mockConfig: Partial = {}): Promise { const ctx = ownContext(new Context()) const { withModelSelection, hostSelectionEnabled = true, parentAgentOptions, ...config } = toolConfig + await mountAgentLoopTestDependencies(ctx) if (withModelSelection === true) { await ctx.plugin(SubagentModelSelectionConfig, { enabled: hostSelectionEnabled, allowedModels: TEST_ALLOWED_MODELS, }) - await mountAgentLoopTestDependencies(ctx) await ctx.plugin(AgentLoop, { agents: [] }) await mountWorkingDirectoryFixture(ctx) await ctx.plugin(SubagentRuntime) @@ -65,9 +66,10 @@ export async function setup(toolConfig: SetupConfig, mockConfig: Partial { + setup: async (agentCtx) => { + await agentCtx.plugin(ListModels) const fiber = agentCtx.inject(tool.inject, (runtimeCtx) => { - tool.apply(runtimeCtx, { ...config, modelSelectionSettings: true }, agent.session) + tool.apply(runtimeCtx, { ...config, modelSelection: true }) }) await fiber.await() }, @@ -75,7 +77,6 @@ export async function setup(toolConfig: SetupConfig, mockConfig: Partial() + async function setupListTool(routes = [ { provider: 'alpha', model: 'fast' }, { provider: 'alpha', model: 'plain' }, { provider: 'beta', model: 'fast' }, { provider: 'beta', model: 'plain' }, ]) { - const ctx = new Context() - await ctx.plugin(LlmRuntime) - await ctx.plugin(SystemPrompt) - await ctx.plugin(ToolRuntime) - registerListSubagentModels(ctx, () => ({ routes })) + const ctx = ownContext(new Context()) + await mountAgentLoopTestDependencies(ctx) + await ctx.plugin(AgentLoop, { agents: [] }) + await ctx.plugin(SubagentModelSelectionConfig, { enabled: true, allowedModels: routes }) + await ctx.plugin(ListModels) + const { agent } = await ctx.agents.create({ sessionId: SessionId('discovery-parent') }) + callingAgents.set(ctx, agent) return ctx } async function setupAllowedListTool() { - const ctx = new Context() - await ctx.plugin(LlmRuntime) - await ctx.plugin(SystemPrompt) - await ctx.plugin(ToolRuntime) - registerListSubagentModels(ctx, () => ({ - routes: [ - { provider: 'alpha', model: 'fast' }, - { provider: 'alpha', model: 'unlisted' }, - { provider: 'missing', model: 'hidden' }, - ], - })) - return ctx + return setupListTool([ + { provider: 'alpha', model: 'fast' }, + { provider: 'alpha', model: 'unlisted' }, + { provider: 'missing', model: 'hidden' }, + ]) } let counter = 0 @@ -98,12 +98,13 @@ function call(ctx: Context, args: unknown, signal = testToolSignal) { callId: ToolCallId(`list-models-${++counter}`), name: 'list_subagent_models', arguments: args, + ...callingAgents.has(ctx) ? { agent: callingAgents.get(ctx)! } : {}, }) } describe('list_subagent_models', () => { - it('is omitted unless its delegation-tool instance owns model selection', async () => { - const ctx = new Context() + it('is mounted independently of delegation tools', async () => { + const ctx = ownContext(new Context()) await ctx.plugin(SessionProjectionRegistry) await ctx.plugin(LlmRuntime) await ctx.plugin(SystemPrompt) @@ -112,23 +113,19 @@ describe('list_subagent_models', () => { await ctx.plugin(SubagentRuntime) await ctx.plugin(tool, { provider: 'unused' }) expect(ctx.tools.get('list_subagent_models')).toBeUndefined() + const discovery = await ctx.plugin(ListModels) + expect(ctx.tools.get('list_subagent_models')).toBeDefined() + await discovery.dispose() + expect(ctx.tools.get('list_subagent_models')).toBeUndefined() + await ctx.fiber.dispose() }) it('requires a calling Agent and an effective parent route', async () => { - const ctx = new Context() - await ctx.plugin(SubagentModelSelectionConfig) + const ctx = ownContext(new Context()) await ctx.plugin(SessionProjectionRegistry) await ctx.plugin(SystemPrompt) await ctx.plugin(ToolRuntime) - await mountWorkingDirectoryFixture(ctx) - await ctx.plugin(SubagentRuntime) - await ctx.plugin({ - name: 'direct-tool-subagent', - inject: tool.inject, - apply: (direct: Context) => { - tool.apply(direct, { provider: 'unused', modelSelectionSettings: true }, Session.create(SessionId('direct-discovery'))) - }, - }) + await ctx.plugin(ListModels) expect(text(await call(ctx, {}))).toContain('requires a calling agent') await ctx.fiber.dispose() @@ -142,7 +139,7 @@ describe('list_subagent_models', () => { expect(text(result)).toContain('without a parent provider and model') }) - it('reads the recorded Session allowlist through the settings-owned discovery definition', async () => { + it('reads the recorded Session allowlist through the independent discovery definition', async () => { const ctx = await setup({ provider: 'mock', withModelSelection: true }) ctx.llm.registerAdapter(['alpha', 'secret'], new CatalogAdapter()) const result = await ctx.tools.execute({ @@ -153,25 +150,20 @@ describe('list_subagent_models', () => { expect(text(result)).toBe('alpha — ALPHA API') }) - it('stays registered without the optional LLM service and rejects discovery calls', async () => { - const ctx = new Context() + it('stays registered without the LLM service and rejects discovery calls', async () => { + const parent = await setupListTool() + const ctx = ownContext(new Context()) + await ctx.plugin(SessionProjectionRegistry) await ctx.plugin(SystemPrompt) await ctx.plugin(ToolRuntime) - await mountWorkingDirectoryFixture(ctx) - await ctx.plugin(SubagentRuntime) - registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'alpha', model: 'fast' }] })) + await ctx.plugin(ListModels) + callingAgents.set(ctx, callingAgents.get(parent)!) + expect(ctx.tools.get('list_subagent_models')).toBeDefined() const result = await call(ctx, {}) expect(result.isError).toBe(true) expect(text(result)).toContain('`llm` service is unavailable') }) - it('rejects two discovery-owning instances in one tool scope', async () => { - const ctx = await setupListTool() - expect(() => { - registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'alpha', model: 'fast' }] })) - }).toThrow('tool "list_subagent_models" is already registered') - }) - it('lists registered providers and follows live registration changes', async () => { const ctx = await setupListTool() const empty = await call(ctx, {}) @@ -309,20 +301,6 @@ describe('list_subagent_models', () => { expect(listModels).not.toHaveBeenCalled() }) - it('requires a calling Agent when a discovery registration supplies no policy', async () => { - const ctx = new Context() - try { - await ctx.plugin(SystemPrompt) - await ctx.plugin(ToolRuntime) - registerListSubagentModels(ctx, () => undefined) - const result = await call(ctx, {}) - expect(result.isError).toBe(true) - expect(text(result)).toContain('requires a calling agent') - } finally { - await ctx.fiber.dispose() - } - }) - it('inspects exact-model efforts, descriptions, and defaults', async () => { const ctx = await setupListTool() ctx.llm.registerAdapter(['alpha', 'secret'], new CatalogAdapter()) diff --git a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts index ff374ed096..975abce19e 100644 --- a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts @@ -4,22 +4,22 @@ import { mountWorkingDirectoryFixture } from '../../subagent/tests/working-direc import { describe, expect, it, vi } from 'vitest' import { Context } from '@deepseek-ai/cordis' import { ToolCallId } from '@deepseek-ai/dsh-llm' -import { Session, SESSION_FORMAT_VERSION, SessionId } from '@deepseek-ai/dsh-session' +import { Session, SessionId } from '@deepseek-ai/dsh-session' import type { SessionEvent } from '@deepseek-ai/dsh-session' import { bindScopeParent, createScope, scopeOf, scopeTarget } from '@deepseek-ai/dsh-scope' import { liveConfig } from '../../../settings/settings/tests/live-config.ts' import AgentLoop from '@deepseek-ai/dsh-agent-loop' import { mountAgentLoopTestDependencies } from '@deepseek-ai/dsh-agent-loop-testkit' -import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection' import SubagentRuntime from '@deepseek-ai/dsh-subagent' import * as SubagentSpawn from '@deepseek-ai/dsh-subagent-spawn-in-process' import * as tool from '../src/index.ts' +import * as ListModels from '../src/list-models.ts' import SubagentModelSelectionConfig from '../src/model-selection-settings.ts' import { subagentModelSelectionPolicy, subagentModelSelectionProjectionDefinition, } from '../src/model-selection-state.ts' -import { callSubagent, text } from './harness.ts' +import { callSubagent, ownContext, text } from './harness.ts' const ALLOWED_MODELS = [{ provider: 'alpha', model: 'fast-model' }] @@ -40,19 +40,20 @@ const modelSelectionPresets = new WeakMap { - const ctx = new Context() + const ctx = ownContext(new Context()) - selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) await mountAgentLoopTestDependencies(ctx) + selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) await ctx.plugin(AgentLoop, { agents: [] }) await mountWorkingDirectoryFixture(ctx) subagentConfigs.set(ctx, await liveConfig(ctx, SubagentRuntime)) await ctx.plugin(SubagentSpawn, { providerName: 'spawn' }) if (withPreset) { const preset = createScope(ctx, { preset: 'model-selection-test' }) + await preset.ctx.plugin(ListModels) await preset.ctx.plugin(tool, { provider: 'spawn', - modelSelectionSettings: true, + modelSelection: true, }) modelSelectionPresets.set(ctx, preset) } @@ -78,7 +79,8 @@ async function createAgent(ctx: Context, id: string, options: { describe('SubagentModelSelectionConfig', () => { it('uses the composed default without a settings provider', async () => { - const ctx = new Context() + const ctx = ownContext(new Context()) + await mountAgentLoopTestDependencies(ctx) await ctx.plugin(SubagentModelSelectionConfig, { enabled: true, allowedModels: ALLOWED_MODELS }) expect(ctx.subagentModelSelection.current()).toEqual({ enabled: true, allowedModels: ALLOWED_MODELS }) @@ -86,7 +88,8 @@ describe('SubagentModelSelectionConfig', () => { }) it('defaults off and follows the validated user layer', async () => { - const ctx = new Context() + const ctx = ownContext(new Context()) + await mountAgentLoopTestDependencies(ctx) selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) @@ -100,10 +103,10 @@ describe('SubagentModelSelectionConfig', () => { }) it('rejects duplicate routes, enabled empty settings, and an empty durable policy', async () => { - const ctx = new Context() + const ctx = ownContext(new Context()) + await mountAgentLoopTestDependencies(ctx) selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) - await ctx.plugin(SessionProjectionRegistry) ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) // A stored policy the service cannot apply fails when it is read, not when the profile row is written. @@ -171,11 +174,12 @@ describe('SubagentModelSelectionConfig', () => { sessionId: SessionId('direct-agent-setup'), setup: async (agentCtx, agent) => { prepared = agent + await agentCtx.plugin(ListModels) const fiber = agentCtx.inject(tool.inject, (runtimeCtx) => { tool.apply(runtimeCtx, { provider: 'spawn', - modelSelectionSettings: true, - }, agent.session) + modelSelection: true, + }) }) await fiber.await() }, @@ -187,7 +191,7 @@ describe('SubagentModelSelectionConfig', () => { await ctx.fiber.dispose() }) - it('captures a blank Session joining the preset without resampling existing Sessions', async () => { + it('preserves creation-time authorization when a blank Session later joins a preset', async () => { const ctx = await boot() const disabled = await createAgent(ctx, 'already-disabled') const joining = await ctx.agents.create({ sessionId: SessionId('joining-preset') }) @@ -195,23 +199,25 @@ describe('SubagentModelSelectionConfig', () => { const preset = modelSelectionPresets.get(ctx)! bindScopeParent(scopeOf(joining.agent.ctx)!, scopeOf(preset.ctx)!) ctx.emit(scopeTarget({}, scopeOf(preset.ctx)), 'tools/change') - expect(subagentModelSelectionPolicy(ctx.sessionProjections, joining.agent.session)).toEqual(ALLOWED_MODELS) + expect(subagentModelSelectionPolicy(ctx.sessionProjections, joining.agent.session)).toBeUndefined() expect(subagentModelSelectionPolicy(ctx.sessionProjections, disabled.session)).toBeUndefined() await ctx.fiber.dispose() }) - it('captures policy for existing Agents only in the mounting preset', async () => { + it('samples authorization without delegation tools and does not resample on tool installation', async () => { const ctx = await boot(false) + await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: ALLOWED_MODELS }) const preset = createScope(ctx, { preset: 'late-tools' }) const inside = await ctx.agents.create({ sessionId: SessionId('existing-inside'), setup: (agentCtx) => { bindScopeParent(scopeOf(agentCtx)!, scopeOf(preset.ctx)!) }, }) const outside = await ctx.agents.create({ sessionId: SessionId('existing-outside') }) - await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: ALLOWED_MODELS }) - await preset.ctx.plugin(tool, { provider: 'spawn', modelSelectionSettings: true }) + await selectionConfigs.get(ctx)!.update({ enabled: false }) + await preset.ctx.plugin(ListModels) + await preset.ctx.plugin(tool, { provider: 'spawn', modelSelection: true }) expect(subagentModelSelectionPolicy(ctx.sessionProjections, inside.agent.session)).toEqual(ALLOWED_MODELS) - expect(subagentModelSelectionPolicy(ctx.sessionProjections, outside.agent.session)).toBeUndefined() + expect(subagentModelSelectionPolicy(ctx.sessionProjections, outside.agent.session)).toEqual(ALLOWED_MODELS) await ctx.fiber.dispose() }) @@ -269,9 +275,10 @@ describe('SubagentModelSelectionConfig', () => { const ctx = await boot(false) const preset = createScope(ctx, { preset: 'standard' }) const other = createScope(ctx, { preset: 'minimal' }) + await preset.ctx.plugin(ListModels) await preset.ctx.plugin(tool, { provider: 'spawn', - modelSelectionSettings: true, + modelSelection: true, }) let enabledBinding: ReturnType | undefined @@ -387,58 +394,18 @@ describe('SubagentModelSelectionConfig', () => { await ctx.fiber.dispose() }) - it('requires both the Host setting owner and a scoped standing preset', async () => { - const withoutSettings = new Context() - await mountAgentLoopTestDependencies(withoutSettings) - await mountWorkingDirectoryFixture(withoutSettings) - await withoutSettings.plugin(SubagentRuntime) - expect(() => { - tool.apply(withoutSettings, { - provider: 'missing', - modelSelectionSettings: true, - maxDepth: 'provider-managed', - }) - }).toThrow('requires @deepseek-ai/dsh-tool-subagent/model-selection-settings') - await withoutSettings.fiber.dispose() - - const withoutAgent = await boot(false) - expect(() => { - tool.apply(withoutAgent, { - provider: 'spawn', - modelSelectionSettings: true, - }) - }).toThrow('requires a scoped preset Context') - - await withoutAgent.fiber.dispose() - }) - - it('requires the Session registry when a child inherits its parent policy', async () => { - const ctx = new Context() - try { - selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) - await ctx.plugin(SessionProjectionRegistry) - await mountWorkingDirectoryFixture(ctx) - subagentConfigs.set(ctx, await liveConfig(ctx, SubagentRuntime)) - const childId = SessionId('child-without-session-registry') - const child = Session.create(childId, undefined, { - version: SESSION_FORMAT_VERSION, - id: childId, - createdAt: 1, - isSeeded: false, - origin: 'subagent', - parentSession: SessionId('missing-parent'), - }) - - expect(() => { - tool.apply(ctx, { - provider: 'missing', - modelSelectionSettings: true, - maxDepth: 'provider-managed', - }, child) - }).toThrow('child model-selection inheritance requires the Session registry') - } finally { - await ctx.fiber.dispose() - } + it('permits independently configured delegation tools without a settings service', async () => { + const ctx = ownContext(new Context()) + await mountAgentLoopTestDependencies(ctx) + await mountWorkingDirectoryFixture(ctx) + await ctx.plugin(SubagentRuntime) + await ctx.plugin(SubagentSpawn, { providerName: 'spawn' }) + await ctx.plugin(tool, { provider: 'spawn', modelSelection: true }) + await ctx.plugin(tool, { provider: 'spawn', toolName: 'another_spawn', modelSelection: true }) + expect(ctx.tools.get('subagent')).toBeDefined() + expect(ctx.tools.get('another_spawn')).toBeDefined() + expect(ctx.tools.get('list_subagent_models')).toBeUndefined() + await ctx.fiber.dispose() }) }) diff --git a/packages/subagent/tool-subagent/tests/model-selection.spec.ts b/packages/subagent/tool-subagent/tests/model-selection.spec.ts index 0181e77b9c..183f1f52f8 100644 --- a/packages/subagent/tool-subagent/tests/model-selection.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection.spec.ts @@ -141,7 +141,7 @@ describe('dsh-tool-subagent model selection', () => { expect(text(result)).toContain('model selection is disabled for this tool instance') }) - it('hides and rejects route fields without modelSelectionSettings', async () => { + it('hides and rejects route fields without modelSelection', async () => { const ctx = await setup({ provider: 'mock', parentAgentOptions: { provider: 'alpha', model: 'parent-model' } }) const schema = ctx.tools.schemas().find(entry => entry.name === 'subagent')! const props = (schema.parameters as { properties?: Record }).properties ?? {} diff --git a/packages/subagent/tool-subagent/tsdown.config.ts b/packages/subagent/tool-subagent/tsdown.config.ts index 4a467f2c5e..8cdb8ec132 100644 --- a/packages/subagent/tool-subagent/tsdown.config.ts +++ b/packages/subagent/tool-subagent/tsdown.config.ts @@ -15,4 +15,5 @@ const entry = (path: string) => ({ export default defineConfig([ entry('lib/types/index.js'), entry('lib/types/model-selection-settings.js'), + entry('lib/types/list-models.js'), ]) diff --git a/scripts/check-workspace-constraints.ts b/scripts/check-workspace-constraints.ts index 1880c55d1e..ce45a292c5 100644 --- a/scripts/check-workspace-constraints.ts +++ b/scripts/check-workspace-constraints.ts @@ -197,7 +197,7 @@ const packageFileExtras: Readonly> = { '@deepseek-ai/dsh-agent-preset': ['skills'], // The Web Host mounts the default-off settings owner independently of the // delegation-tool instances. - '@deepseek-ai/dsh-tool-subagent': ['lib/model-selection-settings.js'], + '@deepseek-ai/dsh-tool-subagent': ['lib/model-selection-settings.js', 'lib/list-models.js'], // The JSONL backend resolves its private verification Worker relative to // import.meta.url; it is shipped without a public package subpath. '@deepseek-ai/dsh-session-persistence-jsonl': ['lib/worker.cjs'], diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index 2d554d3542..937fcb6186 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -73,7 +73,7 @@ import * as PluginManagerTools from '@deepseek-ai/dsh-plugin-manager/tools' import SandboxPolicy from '@deepseek-ai/dsh-sandbox-policy' import McpResources from '@deepseek-ai/dsh-mcp-resources' import * as ToolSubagent from '@deepseek-ai/dsh-tool-subagent' -import { registerListSubagentModels } from '../packages/subagent/tool-subagent/src/list-models.ts' +import * as ListSubagentModels from '../packages/subagent/tool-subagent/src/list-models.ts' import * as ToolWeb from '@deepseek-ai/dsh-tool-web' import WorkflowEngine from '@deepseek-ai/dsh-workflow' import type { WorkflowRun, WorkflowStartRequest } from '@deepseek-ai/dsh-workflow' @@ -519,10 +519,10 @@ const TOOL_PACKAGES: ToolPackage[] = [ await ctx.plugin(LlmRuntime) registerCatalogSubagentProvider(ctx, 'mock') await ctx.plugin(ToolSubagent, { provider: 'mock' }) - registerListSubagentModels(ctx, () => ({ routes: [{ provider: 'mock', model: 'mock' }] })) + await ctx.plugin(ListSubagentModels) }, note: - 'The registered delegation name is the load-time `toolName` config (default `subagent`); the default schema above has model selection off, while the discovery schema is shown as the fixed companion registered by an instance with `modelSelectionSettings`. That instance always exposes optional route fields; Web presets sample the Plugins preference for each new top-level Session and preserve that decision for its child Sessions, and without an enabled policy, selection and discovery are restricted to the current parent route. The shipped `subagent_fork` exposes no route selection. Each instance selects its backend through `provider`. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', + 'The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave fork route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', }, { pkg: '@deepseek-ai/dsh-tool-subagent-control', diff --git a/tsconfig.base.json b/tsconfig.base.json index 0d5ec2c48a..dfb4faf2eb 100644 --- a/tsconfig.base.json +++ b/tsconfig.base.json @@ -185,6 +185,7 @@ "@deepseek-ai/dsh-tools/presentation": ["./packages/core/tools/src/presentation.ts"], "@deepseek-ai/dsh-tools/types": ["./packages/core/tools/src/types.ts"], "@deepseek-ai/dsh-tool-subagent-control/list-agents": ["./packages/subagent/tool-subagent-control/src/list-agents.ts"], + "@deepseek-ai/dsh-tool-subagent/list-models": ["./packages/subagent/tool-subagent/src/list-models.ts"], "@deepseek-ai/dsh-tool-subagent/model-selection-settings": ["./packages/subagent/tool-subagent/src/model-selection-settings.ts"], "@deepseek-ai/dsh-user-approval/types": ["./packages/interaction/user-approval/src/types.ts"], "@deepseek-ai/dsh-user-questions/types": ["./packages/interaction/user-questions/src/types.ts"], From 26d038ae2b3c59797402a0daec9bee7c0e8ee8c0 Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Tue, 13 Oct 2026 11:56:32 +0800 Subject: [PATCH 019/162] fix(subagent): complete discovery plugin metadata and catalog expectations --- .../expected/settings-chrome/plugin-instances.expected.md | 4 ++++ docs/tool-catalog.i18n.yaml | 8 ++++---- docs/tool-catalog.md | 4 ++-- docs/tool-catalog.zh.md | 4 ++-- .../subagent/tool-subagent/locale/list-models/en.json | 5 +++++ .../subagent/tool-subagent/locale/list-models/zh.json | 5 +++++ packages/subagent/tool-subagent/package.json | 4 +++- scripts/gen-tool-catalog.ts | 2 +- 8 files changed, 26 insertions(+), 10 deletions(-) create mode 100644 packages/subagent/tool-subagent/locale/list-models/en.json create mode 100644 packages/subagent/tool-subagent/locale/list-models/zh.json diff --git a/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md b/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md index 239b4855e7..223f7ab367 100644 --- a/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md +++ b/apps/web/tests/expected/settings-chrome/plugin-instances.expected.md @@ -8,6 +8,10 @@ - button "子智能体列表工具, tool-subagent-list-agents, 已启用": - strong: 子智能体列表工具 - code: tool-subagent-list-agents + - listitem: + - button "子智能体模型查询工具, tool-subagent-list-models, 已启用": + - strong: 子智能体模型查询工具 + - code: tool-subagent-list-models - listitem: - button "子智能体委派工具, tool-subagent, 已启用": - strong: 子智能体委派工具 diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 86a316406d..47b26056f4 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: 7224f159dd69abba - zh: 00b69ab3f899010a + en: e4b330276e8729aa + zh: 3f781cb32fd16d81 /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -228,8 +228,8 @@ en: 6872bee11fccb6c8 zh: 230f2d443dc5d9a6 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: - en: 7b06bc34df756d56 - zh: bc82dab293f57b5b + en: 5bcfcf9ca4117096 + zh: 38cb646632b457b2 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent-control: en: ab82e130c3413f04 zh: ab82e130c3413f04 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 7ff20028a7..3b2c922a52 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave fork route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave `subagent_fork` route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2067,7 +2067,7 @@ Delegate a self-contained task to a subagent (a separate agent that works in its Source: [`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave fork route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. +The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave `subagent_fork` route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index 5231594677..fc60890bf1 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 fork 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 `subagent_fork` 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2073,7 +2073,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， 来源：[`packages/subagent/tool-subagent/src/index.ts`](../packages/subagent/tool-subagent/src/index.ts) -加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 fork 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 +加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 `subagent_fork` 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 diff --git a/packages/subagent/tool-subagent/locale/list-models/en.json b/packages/subagent/tool-subagent/locale/list-models/en.json new file mode 100644 index 0000000000..62b37b2526 --- /dev/null +++ b/packages/subagent/tool-subagent/locale/list-models/en.json @@ -0,0 +1,5 @@ +{ + "meta": { + "title": "Subagent model discovery" + } +} diff --git a/packages/subagent/tool-subagent/locale/list-models/zh.json b/packages/subagent/tool-subagent/locale/list-models/zh.json new file mode 100644 index 0000000000..c78ffcc47f --- /dev/null +++ b/packages/subagent/tool-subagent/locale/list-models/zh.json @@ -0,0 +1,5 @@ +{ + "meta": { + "title": "子智能体模型查询工具" + } +} diff --git a/packages/subagent/tool-subagent/package.json b/packages/subagent/tool-subagent/package.json index 00f6896dc0..bfd7789128 100644 --- a/packages/subagent/tool-subagent/package.json +++ b/packages/subagent/tool-subagent/package.json @@ -29,10 +29,12 @@ "./list-models": { "types": "./lib/types/list-models.d.ts", "default": "./lib/list-models.js" - } + }, + "./list-models/locale/*.json": "./locale/list-models/*.json" }, "files": [ "locale/*.json", + "locale/list-models/*.json", "locale/model-selection-settings/*.json", "lib/index.js", "lib/model-selection-settings.js", diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index 937fcb6186..14914a6437 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -522,7 +522,7 @@ const TOOL_PACKAGES: ToolPackage[] = [ await ctx.plugin(ListSubagentModels) }, note: - 'The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave fork route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', + 'The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave `subagent_fork` route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically.', }, { pkg: '@deepseek-ai/dsh-tool-subagent-control', From f12fa3bad186d121e4b649f086354fe09aec6f09 Mon Sep 17 00:00:00 2001 From: yudshj Date: Tue, 13 Oct 2026 10:23:18 +0800 Subject: [PATCH 020/162] fix(office): keep render replies and PDF checks within available files - office-render: report inspection results in text and never link or embed removed previews. Desktop and Web replies otherwise embed deleted preview images, because their file-reference guidance encourages inline images. - office-export-pdf: validate with the command result, nonempty output, and %PDF- signature; do not search for PDF readers unless the user requests deeper checks. The bundled runtime has none, so the old conditional made the model probe system Python and command-line tools. - Assert that the PDF export e2e does not probe PDF reader tools. - Replace stale "structural checks" wording in the Desktop Office plugin JSDoc and the skill package description. --- apps/desktop-host/src/office.ts | 2 +- .../cordis-composition-reference/references/packages.md | 2 +- packages/skill/skill-office/README.i18n.yaml | 4 ++-- packages/skill/skill-office/README.md | 2 +- packages/skill/skill-office/README.zh.md | 2 +- .../skill/skill-office/assets/office-export-pdf/SKILL.md | 2 +- packages/skill/skill-office/assets/office-render/SKILL.md | 2 +- packages/skill/skill-office/package.json | 2 +- packages/skill/skill-office/tests/render-tools.e2e.ts | 1 + .../session/office-progressive-native/session.v4.jsonl | 2 +- snapshots/session/office-progressive-ptc/session.v4.jsonl | 8 ++++---- .../session/office-sidecar-instructions/session.v4.jsonl | 4 ++-- 12 files changed, 17 insertions(+), 16 deletions(-) diff --git a/apps/desktop-host/src/office.ts b/apps/desktop-host/src/office.ts index 13941060ef..7ed14dd3b3 100644 --- a/apps/desktop-host/src/office.ts +++ b/apps/desktop-host/src/office.ts @@ -21,7 +21,7 @@ export interface Config { } /** - * Enable offline Office authoring and structural checks in the Desktop profile. + * Enable offline Office authoring and LibreOffice Kit operations in the Desktop profile. * @param ctx - Profile scope; child plugins declare their own service requirements. * @param config - Bundled payload source and Harness-home installation root. */ diff --git a/packages/preset/agent-preset/skills/cordis-composition-reference/references/packages.md b/packages/preset/agent-preset/skills/cordis-composition-reference/references/packages.md index 8446d4bb43..91df92b78c 100644 --- a/packages/preset/agent-preset/skills/cordis-composition-reference/references/packages.md +++ b/packages/preset/agent-preset/skills/cordis-composition-reference/references/packages.md @@ -403,7 +403,7 @@ Every package below exports a Cordis plugin that a bundle patch can name in a Lo |---|---|---| | `@deepseek-ai/dsh-skill` | yes | Agent skill provider registry for the DeepSeek Harness | | `@deepseek-ai/dsh-skill-filesystem` | yes | Local filesystem skill provider for the DeepSeek Harness | -| `@deepseek-ai/dsh-skill-office` | yes | Bundled Word, PowerPoint, and Excel workflows and structural checks | +| `@deepseek-ai/dsh-skill-office` | yes | Bundled Word, PowerPoint, and Excel workflows and LibreOffice Kit operations | | `@deepseek-ai/dsh-tool-skill` | yes | Model-facing skill loading tool for the DeepSeek Harness | | `@deepseek-ai/dsh-tool-workspace-dependencies` | yes | The load_workspace_dependencies tool: absolute paths into a bundled Python, Node.js, and pnpm payload | diff --git a/packages/skill/skill-office/README.i18n.yaml b/packages/skill/skill-office/README.i18n.yaml index 818c24634e..8be35a0ccb 100644 --- a/packages/skill/skill-office/README.i18n.yaml +++ b/packages/skill/skill-office/README.i18n.yaml @@ -18,8 +18,8 @@ en: bc0460f837676d8a zh: be1692483d7f5daf /deepseek-ai-dsh-skill-office/use-this-package/progressive-loading: - en: b20b58bfaab4d3f6 - zh: 084528c543b8746e + en: d251c551dc229a2b + zh: 82201ee78d37948a /deepseek-ai-dsh-skill-office/understand-the-implementation: en: 6946e30f3663c72f zh: 81dd5de9c7e61562 diff --git a/packages/skill/skill-office/README.md b/packages/skill/skill-office/README.md index 28539b1e4a..12216dbe2f 100644 --- a/packages/skill/skill-office/README.md +++ b/packages/skill/skill-office/README.md @@ -75,7 +75,7 @@ flowchart TD PC --> F ``` -New files and content, data, or formula edits do not require rendering solely because they include incidental basic formatting. Preview images use a private, randomly named temporary directory inside the workspace and are removed after inspection. Loaded instructions remain in session history after an operation finishes. The diagram describes ordinary presets across supported clients; it does not change the shell-only definition of minimal or sdk-minimal. +New files and content, data, or formula edits do not require rendering solely because they include incidental basic formatting. Preview images use a private, randomly named temporary directory inside the workspace and are removed after inspection; replies report the inspection in text without linking them. PDF export validates the command result and `%PDF-` signature without searching for PDF readers. Loaded instructions remain in session history after an operation finishes. The diagram describes ordinary presets across supported clients; it does not change the shell-only definition of minimal or sdk-minimal. ----- diff --git a/packages/skill/skill-office/README.zh.md b/packages/skill/skill-office/README.zh.md index a9d2f5497b..9d464a800e 100644 --- a/packages/skill/skill-office/README.zh.md +++ b/packages/skill/skill-office/README.zh.md @@ -75,7 +75,7 @@ flowchart TD PC --> F ``` -新建文件以及内容、数据或公式修改，不会仅因附带基础格式调整而需要渲染。预览图片保存在工作区内随机命名的私有临时目录中，并在检查后删除。操作结束后，已加载的说明仍属于会话历史。图中描述各端常规预设，不改变 minimal 或 sdk-minimal 的纯 shell 定义。 +新建文件以及内容、数据或公式修改，不会仅因附带基础格式调整而需要渲染。预览图片保存在工作区内随机命名的私有临时目录中，并在检查后删除；回复以文字说明检查结果，不链接这些图片。PDF 导出校验命令结果和 `%PDF-` 签名，不另外查找 PDF 阅读工具。操作结束后，已加载的说明仍属于会话历史。图中描述各端常规预设，不改变 minimal 或 sdk-minimal 的纯 shell 定义。 ----- diff --git a/packages/skill/skill-office/assets/office-export-pdf/SKILL.md b/packages/skill/skill-office/assets/office-export-pdf/SKILL.md index db5d2944af..471610e2f7 100644 --- a/packages/skill/skill-office/assets/office-export-pdf/SKILL.md +++ b/packages/skill/skill-office/assets/office-export-pdf/SKILL.md @@ -13,6 +13,6 @@ Convert the final saved Office file to a new PDF in the task workspace. Preserve "" "" convert --input --output ``` -Check the command result, nonempty output, and `%PDF-` file signature. If a PDF reader library is already available, compare page count and extractable text with the requested content. If it is unavailable, report the extent of validation; do not generate images or install extra tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts. +Check the command result, nonempty output, and `%PDF-` file signature, then report that extent of validation. Unless the user requests deeper checks, do not search for PDF readers, generate images, or install tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts. If export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place. diff --git a/packages/skill/skill-office/assets/office-render/SKILL.md b/packages/skill/skill-office/assets/office-render/SKILL.md index 88e95e59c7..df14704043 100644 --- a/packages/skill/skill-office/assets/office-render/SKILL.md +++ b/packages/skill/skill-office/assets/office-render/SKILL.md @@ -11,7 +11,7 @@ Before generating images only for inspection, establish that the current model a ## Render the saved deliverable -Keep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d "./.dsh-office.XXXXXXXX")` (owner-only permissions) and `--output-dir "$preview_root/render"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString('N')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself. +Keep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d "./.dsh-office.XXXXXXXX")` (owner-only permissions) and `--output-dir "$preview_root/render"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString('N')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself. Report inspection results in text; the reply must not link or embed removed previews. ```text "" "" render --input --output-dir --pages "1,3,4" diff --git a/packages/skill/skill-office/package.json b/packages/skill/skill-office/package.json index c14ba723b9..af60df4389 100644 --- a/packages/skill/skill-office/package.json +++ b/packages/skill/skill-office/package.json @@ -1,6 +1,6 @@ { "name": "@deepseek-ai/dsh-skill-office", - "description": "Bundled Word, PowerPoint, and Excel workflows and structural checks", + "description": "Bundled Word, PowerPoint, and Excel workflows and LibreOffice Kit operations", "version": "0.2.1-alpha.2", "publishConfig": { "access": "public" diff --git a/packages/skill/skill-office/tests/render-tools.e2e.ts b/packages/skill/skill-office/tests/render-tools.e2e.ts index 2ad026253d..f2dc0cfbf2 100644 --- a/packages/skill/skill-office/tests/render-tools.e2e.ts +++ b/packages/skill/skill-office/tests/render-tools.e2e.ts @@ -123,6 +123,7 @@ it.skipIf(!process.env.DEEPSEEK_API_KEY || !runtime)('exports PDF through the op .map(call => (JSON.parse(call.arguments) as { command: string }).command).join('\n') expect(commands).toMatch(/\sconvert\b/u) expect(commands).not.toMatch(/\srender\b/u) + expect(commands).not.toMatch(/pypdf|PyPDF2|\bfitz\b|pdfminer|pdfplumber|pdfinfo|pdftotext/iu) expect((await readFile(join(cwd, 'result.pdf'))).subarray(0, 5).toString()).toBe('%PDF-') expect(await readFile(join(cwd, 'input.docx'))).toEqual(original) }, diff --git a/snapshots/session/office-progressive-native/session.v4.jsonl b/snapshots/session/office-progressive-native/session.v4.jsonl index 14ca061e74..f13f075c4c 100644 --- a/snapshots/session/office-progressive-native/session.v4.jsonl +++ b/snapshots/session/office-progressive-native/session.v4.jsonl @@ -30,7 +30,7 @@ {"type":"step/start","data":{"turn":1,"step":4}} {"type":"assistant/message","data":{"turn":1,"step":4,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_office_4","name":"skill","arguments":"{\"name\":\"office-export-pdf\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:11}}"},"stream":[{"type":"chunk","time":1791559898486,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791559898486,"index":0,"dt":[],"id":"call_office_4","name":"skill","args":["{\"name\":\"office-export-pdf\"}"]},{"type":"chunk","time":1791559898486,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_office_4","name":"skill","arguments":"{\"name\":\"office-export-pdf\"}"}}},{"type":"chunk","time":1791559898486,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":4,"callId":"call_office_4","name":"skill","arguments":"{\"name\":\"office-export-pdf\"}"}} -{"type":"tool/result","data":{"turn":1,"step":4,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_4"},"toolCallId":"call_office_4","content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-export-pdf\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Export an Office document to PDF\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\n\nConvert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\n\n```text\n\"\" \"\" convert --input --output \n```\n\nCheck the command result, nonempty output, and `%PDF-` file signature. If a PDF reader library is already available, compare page count and extractable text with the requested content. If it is unavailable, report the extent of validation; do not generate images or install extra tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\n\nIf export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\n\n## Installed LibreOffice Kit\n\nUse these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\n\n{\n \"libreofficeKit\": {\n \"node\": \"{{cwd}}/office-node\",\n \"cli\": \"{{cwd}}/render-cli.cjs\"\n }\n}\n\n"}],"isError":false,"id":"{{message:12}}"}},"sourceEventSeqs":[30],"surfaceOp":"append"} +{"type":"tool/result","data":{"turn":1,"step":4,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_4"},"toolCallId":"call_office_4","content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-export-pdf\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Export an Office document to PDF\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\n\nConvert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\n\n```text\n\"\" \"\" convert --input --output \n```\n\nCheck the command result, nonempty output, and `%PDF-` file signature, then report that extent of validation. Unless the user requests deeper checks, do not search for PDF readers, generate images, or install tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\n\nIf export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\n\n## Installed LibreOffice Kit\n\nUse these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\n\n{\n \"libreofficeKit\": {\n \"node\": \"{{cwd}}/office-node\",\n \"cli\": \"{{cwd}}/render-cli.cjs\"\n }\n}\n\n"}],"isError":false,"id":"{{message:12}}"}},"sourceEventSeqs":[30],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":4}} {"type":"step/start","data":{"turn":1,"step":5}} {"type":"assistant/message","data":{"turn":1,"step":5,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:13}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1791559898502,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1791559898502,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1791559898502,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1791559898502,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1791559898502,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} diff --git a/snapshots/session/office-progressive-ptc/session.v4.jsonl b/snapshots/session/office-progressive-ptc/session.v4.jsonl index 9bd58408bb..54f95fab62 100644 --- a/snapshots/session/office-progressive-ptc/session.v4.jsonl +++ b/snapshots/session/office-progressive-ptc/session.v4.jsonl @@ -30,15 +30,15 @@ {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_office_3","name":"run_code","arguments":"{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-render\\\"}));\",\"description\":\"Load task instructions\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:9}}"},"stream":[{"type":"chunk","time":1791781292389,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791781292389,"index":0,"dt":[],"id":"call_office_3","name":"run_code","args":["{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-render\\\"}));\",\"description\":\"Load task instructions\"}"]},{"type":"chunk","time":1791781292389,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_office_3","name":"run_code","arguments":"{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-render\\\"}));\",\"description\":\"Load task instructions\"}"}}},{"type":"chunk","time":1791781292389,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":3,"callId":"call_office_3","name":"run_code","arguments":"{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-render\\\"}));\",\"description\":\"Load task instructions\"}"}} {"type":"tool/ptc-dispatch-start","data":{"rootCallId":"call_office_3","parentCallId":"call_office_3","subCallId":"call_office_3:ptc:1","name":"skill","arguments":{"name":"office-render"}}} -{"type":"tool/ptc-dispatch","data":{"rootCallId":"call_office_3","parentCallId":"call_office_3","subCallId":"call_office_3:ptc:1","name":"skill","arguments":{"name":"office-render"},"isError":false,"content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-render\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Document images and visual inspection\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, install a replacement renderer, or build an alternative rendering pipeline to work around a failed render with an installed sidecar.\n\nBefore generating images only for inspection, establish that the current model accepts images. Otherwise finish the applicable content and structure checks and report the requested visual check as unavailable. Do not generate unreadable previews, retry image reads, or choose a second model. Requested image exports can still be delivered.\n\n## Render the saved deliverable\n\nKeep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d \"./.dsh-office.XXXXXXXX\")` (owner-only permissions) and `--output-dir \"$preview_root/render\"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString('N')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself.\n\n```text\n\"\" \"\" render --input --output-dir --pages \"1,3,4\"\n\"\" \"\" render --input --output-dir --sheet --range \n```\n\nFor DOCX, PPTX, or PDF, batch requested pages/slides in one call. In PowerShell, quote comma-separated `--pages` values so they remain one argument. Inspect all pages only for a requested whole-document review. For worksheets, choose the exact affected range, including chart/drawing regions; worksheet numbers are not PDF page numbers. Inspect the final saved file, reuse existing images for an unchanged source, and rerender only after requested source changes or for previously unrendered regions. Change print areas or scaling only when the user requests print layout.\n\nRead the returned JSON manifest for image paths, page counts, fragment rectangles, and missing-font diagnostics. Use the session's image-reading tool to inspect the requested images. Check clipping, alignment, headings, table widths, contrast, chart labels, and number formats as relevant to the task. A successful command or image read alone does not establish a usable preview; guessed text geometry does not establish overflow.\n\n## Failed or blank previews\n\nFor populated content, a fully transparent or uniformly blank image is a renderer failure. If a pixel check is needed, composite RGBA onto white first; raw RGB under transparent pixels can look nonblank. Check once and do not reread the same image. Stop visual checks for the task immediately after a blank result. This overrides the rerender-after-edit guidance above: do not render or read images again, even after editing or copying the source, or try another range, page, sheet, DPI, output format, diagnostic export, or print setting. Preserve the usable source and requested outputs and finish applicable content/data checks.\n\nIf content and structure checks pass, name the saved file and verified edits as needed, then include exactly one visual-status sentence: \"Content and structure checks passed; visual inspection was not completed.\" in English, or \"内容与结构检查通过；视觉检查未完成。\" in Chinese. Do not add a render or preview bullet, explain why inspection stopped, or describe the image or pixels. Report content or structure failures separately; never say all checks passed or claim visual verification.\n\nLibreOffice output can differ from Word, PowerPoint, or Excel. Images do not certify animation/media playback, refresh formula caches in the source workbook, or guarantee native application fidelity. Deliver requested files through `present` when available; keep the files in place. Do not deliver inspection images or reports unless requested.\n\n## Installed LibreOffice Kit\n\nUse these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\n\n{\n \"libreofficeKit\": {\n \"node\": \"{{cwd}}/office-node\",\n \"cli\": \"{{cwd}}/render-cli.cjs\"\n }\n}\n\n"}]}} -{"type":"tool/result","data":{"turn":1,"step":3,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_3"},"toolCallId":"call_office_3","content":[{"type":"text","text":"{\n name: 'office-render',\n provider: 'dsh-office',\n resourceBase: {\n kind: 'directory',\n path: '{{cwd}}/office-skills/office-render'\n },\n content: '# Document images and visual inspection\\n' +\n '\\n' +\n 'Use the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, install a replacement renderer, or build an alternative rendering pipeline to work around a failed render with an installed sidecar.\\n' +\n '\\n' +\n 'Before generating images only for inspection, establish that the current model accepts images. Otherwise finish the applicable content and structure checks and report the requested visual check as unavailable. Do not generate unreadable previews, retry image reads, or choose a second model. Requested image exports can still be delivered.\\n' +\n '\\n' +\n '## Render the saved deliverable\\n' +\n '\\n' +\n 'Keep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d \"./.dsh-office.XXXXXXXX\")` (owner-only permissions) and `--output-dir \"$preview_root/render\"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString(\\'N\\')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself.\\n' +\n '\\n' +\n '```text\\n' +\n '\"\" \"\" render --input --output-dir --pages \"1,3,4\"\\n' +\n '\"\" \"\" render --input --output-dir --sheet --range \\n' +\n '```\\n' +\n '\\n' +\n 'For DOCX, PPTX, or PDF, batch requested pages/slides in one call. In PowerShell, quote comma-separated `--pages` values so they remain one argument. Inspect all pages only for a requested whole-document review. For worksheets, choose the exact affected range, including chart/drawing regions; worksheet numbers are not PDF page numbers. Inspect the final saved file, reuse existing images for an unchanged source, and rerender only after requested source changes or for previously unrendered regions. Change print areas or scaling only when the user requests print layout.\\n' +\n '\\n' +\n \"Read the returned JSON manifest for image paths, page counts, fragment rectangles, and missing-font diagnostics. Use the session's image-reading tool to inspect the requested images. Check clipping, alignment, headings, table widths, contrast, chart labels, and number formats as relevant to the task. A successful command or image read alone does not establish a usable preview; guessed text geometry does not establish overflow.\\n\" +\n '\\n' +\n '## Failed or blank previews\\n' +\n '\\n' +\n 'For populated content, a fully transparent or uniformly blank image is a renderer failure. If a pixel check is needed, composite RGBA onto white first; raw RGB under transparent pixels can look nonblank. Check once and do not reread the same image. Stop visual checks for the task immediately after a blank result. This overrides the rerender-after-edit guidance above: do not render or read images again, even after editing or copying the source, or try another range, page, sheet, DPI, output format, diagnostic export, or print setting. Preserve the usable source and requested outputs and finish applicable content/data checks.\\n' +\n '\\n' +\n 'If content and structure checks pass, name the saved file and verified edits as needed, then include exactly one visual-status sentence: \"Content and structure checks passed; visual inspection was not completed.\" in English, or \"内容与结构检查通过；视觉检查未完成。\" in Chinese. Do not add a render or preview bullet, explain why inspection stopped, or describe the image or pixels. Report content or structure failures separately; never say all checks passed or claim visual verification.\\n' +\n '\\n' +\n 'LibreOffice output can differ from Word, PowerPoint, or Excel. Images do not certify animation/media playback, refresh formula caches in the source workbook, or guarantee native application fidelity. Deliver requested files through `present` when available; keep the files in place. Do not deliver inspection images or reports unless requested.\\n' +\n '\\n' +\n '## Installed LibreOffice Kit\\n' +\n '\\n' +\n 'Use these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\\n' +\n '\\n' +\n '{\\n' +\n ' \"libreofficeKit\": {\\n' +\n ' \"node\": \"{{cwd}}/office-node\",\\n' +\n ' \"cli\": \"{{cwd}}/render-cli.cjs\"\\n' +\n ' }\\n' +\n '}'\n}"}],"isError":false,"id":"{{message:10}}"}},"sourceEventSeqs":[29],"surfaceOp":"append"} +{"type":"tool/ptc-dispatch","data":{"rootCallId":"call_office_3","parentCallId":"call_office_3","subCallId":"call_office_3:ptc:1","name":"skill","arguments":{"name":"office-render"},"isError":false,"content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-render\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Document images and visual inspection\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, install a replacement renderer, or build an alternative rendering pipeline to work around a failed render with an installed sidecar.\n\nBefore generating images only for inspection, establish that the current model accepts images. Otherwise finish the applicable content and structure checks and report the requested visual check as unavailable. Do not generate unreadable previews, retry image reads, or choose a second model. Requested image exports can still be delivered.\n\n## Render the saved deliverable\n\nKeep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d \"./.dsh-office.XXXXXXXX\")` (owner-only permissions) and `--output-dir \"$preview_root/render\"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString('N')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself. Report inspection results in text; the reply must not link or embed removed previews.\n\n```text\n\"\" \"\" render --input --output-dir --pages \"1,3,4\"\n\"\" \"\" render --input --output-dir --sheet --range \n```\n\nFor DOCX, PPTX, or PDF, batch requested pages/slides in one call. In PowerShell, quote comma-separated `--pages` values so they remain one argument. Inspect all pages only for a requested whole-document review. For worksheets, choose the exact affected range, including chart/drawing regions; worksheet numbers are not PDF page numbers. Inspect the final saved file, reuse existing images for an unchanged source, and rerender only after requested source changes or for previously unrendered regions. Change print areas or scaling only when the user requests print layout.\n\nRead the returned JSON manifest for image paths, page counts, fragment rectangles, and missing-font diagnostics. Use the session's image-reading tool to inspect the requested images. Check clipping, alignment, headings, table widths, contrast, chart labels, and number formats as relevant to the task. A successful command or image read alone does not establish a usable preview; guessed text geometry does not establish overflow.\n\n## Failed or blank previews\n\nFor populated content, a fully transparent or uniformly blank image is a renderer failure. If a pixel check is needed, composite RGBA onto white first; raw RGB under transparent pixels can look nonblank. Check once and do not reread the same image. Stop visual checks for the task immediately after a blank result. This overrides the rerender-after-edit guidance above: do not render or read images again, even after editing or copying the source, or try another range, page, sheet, DPI, output format, diagnostic export, or print setting. Preserve the usable source and requested outputs and finish applicable content/data checks.\n\nIf content and structure checks pass, name the saved file and verified edits as needed, then include exactly one visual-status sentence: \"Content and structure checks passed; visual inspection was not completed.\" in English, or \"内容与结构检查通过；视觉检查未完成。\" in Chinese. Do not add a render or preview bullet, explain why inspection stopped, or describe the image or pixels. Report content or structure failures separately; never say all checks passed or claim visual verification.\n\nLibreOffice output can differ from Word, PowerPoint, or Excel. Images do not certify animation/media playback, refresh formula caches in the source workbook, or guarantee native application fidelity. Deliver requested files through `present` when available; keep the files in place. Do not deliver inspection images or reports unless requested.\n\n## Installed LibreOffice Kit\n\nUse these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\n\n{\n \"libreofficeKit\": {\n \"node\": \"{{cwd}}/office-node\",\n \"cli\": \"{{cwd}}/render-cli.cjs\"\n }\n}\n\n"}]}} +{"type":"tool/result","data":{"turn":1,"step":3,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_3"},"toolCallId":"call_office_3","content":[{"type":"text","text":"{\n name: 'office-render',\n provider: 'dsh-office',\n resourceBase: {\n kind: 'directory',\n path: '{{cwd}}/office-skills/office-render'\n },\n content: '# Document images and visual inspection\\n' +\n '\\n' +\n 'Use the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, install a replacement renderer, or build an alternative rendering pipeline to work around a failed render with an installed sidecar.\\n' +\n '\\n' +\n 'Before generating images only for inspection, establish that the current model accepts images. Otherwise finish the applicable content and structure checks and report the requested visual check as unavailable. Do not generate unreadable previews, retry image reads, or choose a second model. Requested image exports can still be delivered.\\n' +\n '\\n' +\n '## Render the saved deliverable\\n' +\n '\\n' +\n 'Keep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d \"./.dsh-office.XXXXXXXX\")` (owner-only permissions) and `--output-dir \"$preview_root/render\"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString(\\'N\\')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself. Report inspection results in text; the reply must not link or embed removed previews.\\n' +\n '\\n' +\n '```text\\n' +\n '\"\" \"\" render --input --output-dir --pages \"1,3,4\"\\n' +\n '\"\" \"\" render --input --output-dir --sheet --range \\n' +\n '```\\n' +\n '\\n' +\n 'For DOCX, PPTX, or PDF, batch requested pages/slides in one call. In PowerShell, quote comma-separated `--pages` values so they remain one argument. Inspect all pages only for a requested whole-document review. For worksheets, choose the exact affected range, including chart/drawing regions; worksheet numbers are not PDF page numbers. Inspect the final saved file, reuse existing images for an unchanged source, and rerender only after requested source changes or for previously unrendered regions. Change print areas or scaling only when the user requests print layout.\\n' +\n '\\n' +\n \"Read the returned JSON manifest for image paths, page counts, fragment rectangles, and missing-font diagnostics. Use the session's image-reading tool to inspect the requested images. Check clipping, alignment, headings, table widths, contrast, chart labels, and number formats as relevant to the task. A successful command or image read alone does not establish a usable preview; guessed text geometry does not establish overflow.\\n\" +\n '\\n' +\n '## Failed or blank previews\\n' +\n '\\n' +\n 'For populated content, a fully transparent or uniformly blank image is a renderer failure. If a pixel check is needed, composite RGBA onto white first; raw RGB under transparent pixels can look nonblank. Check once and do not reread the same image. Stop visual checks for the task immediately after a blank result. This overrides the rerender-after-edit guidance above: do not render or read images again, even after editing or copying the source, or try another range, page, sheet, DPI, output format, diagnostic export, or print setting. Preserve the usable source and requested outputs and finish applicable content/data checks.\\n' +\n '\\n' +\n 'If content and structure checks pass, name the saved file and verified edits as needed, then include exactly one visual-status sentence: \"Content and structure checks passed; visual inspection was not completed.\" in English, or \"内容与结构检查通过；视觉检查未完成。\" in Chinese. Do not add a render or preview bullet, explain why inspection stopped, or describe the image or pixels. Report content or structure failures separately; never say all checks passed or claim visual verification.\\n' +\n '\\n' +\n 'LibreOffice output can differ from Word, PowerPoint, or Excel. Images do not certify animation/media playback, refresh formula caches in the source workbook, or guarantee native application fidelity. Deliver requested files through `present` when available; keep the files in place. Do not deliver inspection images or reports unless requested.\\n' +\n '\\n' +\n '## Installed LibreOffice Kit\\n' +\n '\\n' +\n 'Use these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\\n' +\n '\\n' +\n '{\\n' +\n ' \"libreofficeKit\": {\\n' +\n ' \"node\": \"{{cwd}}/office-node\",\\n' +\n ' \"cli\": \"{{cwd}}/render-cli.cjs\"\\n' +\n ' }\\n' +\n '}'\n}"}],"isError":false,"id":"{{message:10}}"}},"sourceEventSeqs":[29],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":3}} {"type":"step/start","data":{"turn":1,"step":4}} {"type":"assistant/message","data":{"turn":1,"step":4,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_office_4","name":"run_code","arguments":"{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-export-pdf\\\"}));\",\"description\":\"Load task instructions\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:11}}"},"stream":[{"type":"chunk","time":1791781292502,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791781292502,"index":0,"dt":[],"id":"call_office_4","name":"run_code","args":["{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-export-pdf\\\"}));\",\"description\":\"Load task instructions\"}"]},{"type":"chunk","time":1791781292502,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_office_4","name":"run_code","arguments":"{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-export-pdf\\\"}));\",\"description\":\"Load task instructions\"}"}}},{"type":"chunk","time":1791781292502,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":4,"callId":"call_office_4","name":"run_code","arguments":"{\"code\":\"console.log(await tools.skill({\\\"name\\\":\\\"office-export-pdf\\\"}));\",\"description\":\"Load task instructions\"}"}} {"type":"tool/ptc-dispatch-start","data":{"rootCallId":"call_office_4","parentCallId":"call_office_4","subCallId":"call_office_4:ptc:1","name":"skill","arguments":{"name":"office-export-pdf"}}} -{"type":"tool/ptc-dispatch","data":{"rootCallId":"call_office_4","parentCallId":"call_office_4","subCallId":"call_office_4:ptc:1","name":"skill","arguments":{"name":"office-export-pdf"},"isError":false,"content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-export-pdf\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Export an Office document to PDF\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\n\nConvert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\n\n```text\n\"\" \"\" convert --input --output \n```\n\nCheck the command result, nonempty output, and `%PDF-` file signature. If a PDF reader library is already available, compare page count and extractable text with the requested content. If it is unavailable, report the extent of validation; do not generate images or install extra tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\n\nIf export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\n\n## Installed LibreOffice Kit\n\nUse these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\n\n{\n \"libreofficeKit\": {\n \"node\": \"{{cwd}}/office-node\",\n \"cli\": \"{{cwd}}/render-cli.cjs\"\n }\n}\n\n"}]}} -{"type":"tool/result","data":{"turn":1,"step":4,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_4"},"toolCallId":"call_office_4","content":[{"type":"text","text":"{\n name: 'office-export-pdf',\n provider: 'dsh-office',\n resourceBase: {\n kind: 'directory',\n path: '{{cwd}}/office-skills/office-export-pdf'\n },\n content: '# Export an Office document to PDF\\n' +\n '\\n' +\n 'Use the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\\n' +\n '\\n' +\n 'Convert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\\n' +\n '\\n' +\n '```text\\n' +\n '\"\" \"\" convert --input --output \\n' +\n '```\\n' +\n '\\n' +\n \"Check the command result, nonempty output, and `%PDF-` file signature. If a PDF reader library is already available, compare page count and extractable text with the requested content. If it is unavailable, report the extent of validation; do not generate images or install extra tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\\n\" +\n '\\n' +\n 'If export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\\n' +\n '\\n' +\n '## Installed LibreOffice Kit\\n' +\n '\\n' +\n 'Use these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\\n' +\n '\\n' +\n '{\\n' +\n ' \"libreofficeKit\": {\\n' +\n ' \"node\": \"{{cwd}}/office-node\",\\n' +\n ' \"cli\": \"{{cwd}}/render-cli.cjs\"\\n' +\n ' }\\n' +\n '}'\n}"}],"isError":false,"id":"{{message:12}}"}},"sourceEventSeqs":[36],"surfaceOp":"append"} +{"type":"tool/ptc-dispatch","data":{"rootCallId":"call_office_4","parentCallId":"call_office_4","subCallId":"call_office_4:ptc:1","name":"skill","arguments":{"name":"office-export-pdf"},"isError":false,"content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-export-pdf\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Export an Office document to PDF\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\n\nConvert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\n\n```text\n\"\" \"\" convert --input --output \n```\n\nCheck the command result, nonempty output, and `%PDF-` file signature, then report that extent of validation. Unless the user requests deeper checks, do not search for PDF readers, generate images, or install tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\n\nIf export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\n\n## Installed LibreOffice Kit\n\nUse these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\n\n{\n \"libreofficeKit\": {\n \"node\": \"{{cwd}}/office-node\",\n \"cli\": \"{{cwd}}/render-cli.cjs\"\n }\n}\n\n"}]}} +{"type":"tool/result","data":{"turn":1,"step":4,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_4"},"toolCallId":"call_office_4","content":[{"type":"text","text":"{\n name: 'office-export-pdf',\n provider: 'dsh-office',\n resourceBase: {\n kind: 'directory',\n path: '{{cwd}}/office-skills/office-export-pdf'\n },\n content: '# Export an Office document to PDF\\n' +\n '\\n' +\n 'Use the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\\n' +\n '\\n' +\n 'Convert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\\n' +\n '\\n' +\n '```text\\n' +\n '\"\" \"\" convert --input --output \\n' +\n '```\\n' +\n '\\n' +\n \"Check the command result, nonempty output, and `%PDF-` file signature, then report that extent of validation. Unless the user requests deeper checks, do not search for PDF readers, generate images, or install tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\\n\" +\n '\\n' +\n 'If export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\\n' +\n '\\n' +\n '## Installed LibreOffice Kit\\n' +\n '\\n' +\n 'Use these absolute paths for every LibreOffice Kit command. Pass the CLI entry as the first argument to Node.\\n' +\n '\\n' +\n '{\\n' +\n ' \"libreofficeKit\": {\\n' +\n ' \"node\": \"{{cwd}}/office-node\",\\n' +\n ' \"cli\": \"{{cwd}}/render-cli.cjs\"\\n' +\n ' }\\n' +\n '}'\n}"}],"isError":false,"id":"{{message:12}}"}},"sourceEventSeqs":[36],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":4}} {"type":"step/start","data":{"turn":1,"step":5}} {"type":"assistant/message","data":{"turn":1,"step":5,"message":{"role":"assistant","content":[{"type":"text","text":"DONE"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:13}}"},"usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10},"stream":[{"type":"chunk","time":1791781292618,"chunk":{"type":"block-start","index":0,"blockType":"text"}},{"type":"text-chunks","time0":1791781292618,"index":0,"dt":[],"texts":["DONE"]},{"type":"chunk","time":1791781292618,"chunk":{"type":"block-end","index":0,"block":{"type":"text","text":"DONE"}}},{"type":"chunk","time":1791781292618,"chunk":{"type":"usage","usage":{"inputTokens":12,"outputTokens":2,"cacheReadTokens":10}}},{"type":"chunk","time":1791781292618,"chunk":{"type":"finish","reason":{"kind":"stop"}}}]},"surfaceOp":"append"} diff --git a/snapshots/session/office-sidecar-instructions/session.v4.jsonl b/snapshots/session/office-sidecar-instructions/session.v4.jsonl index ada93b0337..5b7cbcb94b 100644 --- a/snapshots/session/office-sidecar-instructions/session.v4.jsonl +++ b/snapshots/session/office-sidecar-instructions/session.v4.jsonl @@ -15,12 +15,12 @@ {"type":"session/title","data":{"title":"Load the Office rendering, PDF","messageSeqs":[8],"source":{"kind":"fallback"}}} {"type":"assistant/message","data":{"turn":1,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_office_skill","name":"skill","arguments":"{\"name\":\"office-render\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:5}}"},"stream":[{"type":"chunk","time":1790753571607,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790753571608,"index":0,"dt":[],"id":"call_office_skill","name":"skill","args":["{\"name\":\"office-render\"}"]},{"type":"chunk","time":1790753571608,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_office_skill","name":"skill","arguments":"{\"name\":\"office-render\"}"}}},{"type":"chunk","time":1790753571608,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":1,"callId":"call_office_skill","name":"skill","arguments":"{\"name\":\"office-render\"}"}} -{"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_skill"},"toolCallId":"call_office_skill","content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-render\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Document images and visual inspection\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, install a replacement renderer, or build an alternative rendering pipeline to work around a failed render with an installed sidecar.\n\nBefore generating images only for inspection, establish that the current model accepts images. Otherwise finish the applicable content and structure checks and report the requested visual check as unavailable. Do not generate unreadable previews, retry image reads, or choose a second model. Requested image exports can still be delivered.\n\n## Render the saved deliverable\n\nKeep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d \"./.dsh-office.XXXXXXXX\")` (owner-only permissions) and `--output-dir \"$preview_root/render\"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString('N')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself.\n\n```text\n\"\" \"\" render --input --output-dir --pages \"1,3,4\"\n\"\" \"\" render --input --output-dir --sheet --range \n```\n\nFor DOCX, PPTX, or PDF, batch requested pages/slides in one call. In PowerShell, quote comma-separated `--pages` values so they remain one argument. Inspect all pages only for a requested whole-document review. For worksheets, choose the exact affected range, including chart/drawing regions; worksheet numbers are not PDF page numbers. Inspect the final saved file, reuse existing images for an unchanged source, and rerender only after requested source changes or for previously unrendered regions. Change print areas or scaling only when the user requests print layout.\n\nRead the returned JSON manifest for image paths, page counts, fragment rectangles, and missing-font diagnostics. Use the session's image-reading tool to inspect the requested images. Check clipping, alignment, headings, table widths, contrast, chart labels, and number formats as relevant to the task. A successful command or image read alone does not establish a usable preview; guessed text geometry does not establish overflow.\n\n## Failed or blank previews\n\nFor populated content, a fully transparent or uniformly blank image is a renderer failure. If a pixel check is needed, composite RGBA onto white first; raw RGB under transparent pixels can look nonblank. Check once and do not reread the same image. Stop visual checks for the task immediately after a blank result. This overrides the rerender-after-edit guidance above: do not render or read images again, even after editing or copying the source, or try another range, page, sheet, DPI, output format, diagnostic export, or print setting. Preserve the usable source and requested outputs and finish applicable content/data checks.\n\nIf content and structure checks pass, name the saved file and verified edits as needed, then include exactly one visual-status sentence: \"Content and structure checks passed; visual inspection was not completed.\" in English, or \"内容与结构检查通过；视觉检查未完成。\" in Chinese. Do not add a render or preview bullet, explain why inspection stopped, or describe the image or pixels. Report content or structure failures separately; never say all checks passed or claim visual verification.\n\nLibreOffice output can differ from Word, PowerPoint, or Excel. Images do not certify animation/media playback, refresh formula caches in the source workbook, or guarantee native application fidelity. Deliver requested files through `present` when available; keep the files in place. Do not deliver inspection images or reports unless requested.\n\nLibreOffice Kit is disabled in this deployment.\n\n"}],"isError":false,"id":"{{message:6}}"}},"sourceEventSeqs":[15],"surfaceOp":"append"} +{"type":"tool/result","data":{"turn":1,"step":1,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_skill"},"toolCallId":"call_office_skill","content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-render\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Document images and visual inspection\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, install a replacement renderer, or build an alternative rendering pipeline to work around a failed render with an installed sidecar.\n\nBefore generating images only for inspection, establish that the current model accepts images. Otherwise finish the applicable content and structure checks and report the requested visual check as unavailable. Do not generate unreadable previews, retry image reads, or choose a second model. Requested image exports can still be delivered.\n\n## Render the saved deliverable\n\nKeep the saved deliverable in the task workspace. For requested image exports, use the requested output directory and retain the images. For inspection-only previews, create a private, uniquely named temporary parent inside that workspace and pass a child path that does not yet exist as `--output-dir`; the CLI creates that child. On POSIX, use `preview_root=$(mktemp -d \"./.dsh-office.XXXXXXXX\")` (owner-only permissions) and `--output-dir \"$preview_root/render\"`. On PowerShell, create a parent named `.dsh-office-` plus `[guid]::NewGuid().ToString('N')` under the task workspace and pass an uncreated `render` child; it inherits the workspace ACL. Quote each input and output path. Retain the parent until image inspection finishes, then remove only that parent, including after a failed render. Never use a fixed directory name or remove the workspace itself. Report inspection results in text; the reply must not link or embed removed previews.\n\n```text\n\"\" \"\" render --input --output-dir --pages \"1,3,4\"\n\"\" \"\" render --input --output-dir --sheet --range \n```\n\nFor DOCX, PPTX, or PDF, batch requested pages/slides in one call. In PowerShell, quote comma-separated `--pages` values so they remain one argument. Inspect all pages only for a requested whole-document review. For worksheets, choose the exact affected range, including chart/drawing regions; worksheet numbers are not PDF page numbers. Inspect the final saved file, reuse existing images for an unchanged source, and rerender only after requested source changes or for previously unrendered regions. Change print areas or scaling only when the user requests print layout.\n\nRead the returned JSON manifest for image paths, page counts, fragment rectangles, and missing-font diagnostics. Use the session's image-reading tool to inspect the requested images. Check clipping, alignment, headings, table widths, contrast, chart labels, and number formats as relevant to the task. A successful command or image read alone does not establish a usable preview; guessed text geometry does not establish overflow.\n\n## Failed or blank previews\n\nFor populated content, a fully transparent or uniformly blank image is a renderer failure. If a pixel check is needed, composite RGBA onto white first; raw RGB under transparent pixels can look nonblank. Check once and do not reread the same image. Stop visual checks for the task immediately after a blank result. This overrides the rerender-after-edit guidance above: do not render or read images again, even after editing or copying the source, or try another range, page, sheet, DPI, output format, diagnostic export, or print setting. Preserve the usable source and requested outputs and finish applicable content/data checks.\n\nIf content and structure checks pass, name the saved file and verified edits as needed, then include exactly one visual-status sentence: \"Content and structure checks passed; visual inspection was not completed.\" in English, or \"内容与结构检查通过；视觉检查未完成。\" in Chinese. Do not add a render or preview bullet, explain why inspection stopped, or describe the image or pixels. Report content or structure failures separately; never say all checks passed or claim visual verification.\n\nLibreOffice output can differ from Word, PowerPoint, or Excel. Images do not certify animation/media playback, refresh formula caches in the source workbook, or guarantee native application fidelity. Deliver requested files through `present` when available; keep the files in place. Do not deliver inspection images or reports unless requested.\n\nLibreOffice Kit is disabled in this deployment.\n\n"}],"isError":false,"id":"{{message:6}}"}},"sourceEventSeqs":[15],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":1}} {"type":"step/start","data":{"turn":1,"step":2}} {"type":"assistant/message","data":{"turn":1,"step":2,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_office_2","name":"skill","arguments":"{\"name\":\"office-export-pdf\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:7}}"},"stream":[{"type":"chunk","time":1790753571639,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790753571639,"index":0,"dt":[],"id":"call_office_2","name":"skill","args":["{\"name\":\"office-export-pdf\"}"]},{"type":"chunk","time":1790753571639,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_office_2","name":"skill","arguments":"{\"name\":\"office-export-pdf\"}"}}},{"type":"chunk","time":1790753571639,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":1,"step":2,"callId":"call_office_2","name":"skill","arguments":"{\"name\":\"office-export-pdf\"}"}} -{"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_2"},"toolCallId":"call_office_2","content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-export-pdf\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Export an Office document to PDF\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\n\nConvert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\n\n```text\n\"\" \"\" convert --input --output \n```\n\nCheck the command result, nonempty output, and `%PDF-` file signature. If a PDF reader library is already available, compare page count and extractable text with the requested content. If it is unavailable, report the extent of validation; do not generate images or install extra tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\n\nIf export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\n\nLibreOffice Kit is disabled in this deployment.\n\n"}],"isError":false,"id":"{{message:8}}"}},"sourceEventSeqs":[20],"surfaceOp":"append"} +{"type":"tool/result","data":{"turn":1,"step":2,"message":{"role":"tool","source":{"kind":"tool","callId":"call_office_2"},"toolCallId":"call_office_2","content":[{"type":"text","text":"\n\nBase directory for this skill: {{cwd}}/office-skills/office-export-pdf\nResolve relative paths mentioned by this skill against the base directory before using them. Load referenced resources only as needed.\n\n\n\n# Export an Office document to PDF\n\nUse the installed LibreOffice Kit CLI through the shell with the absolute Node and CLI paths supplied below. In Bash, invoke `\"\" \"\"`; in PowerShell, invoke `& \"\" \"\"`. Prefer bundled LibreOffice unless the user requests another installation. If the Office sidecar is missing, follow [sidecar setup](../office/references/sidecar.md): ask whether the user wants to download it and install only after they agree. Do not invoke system soffice, use npx, or install a replacement to work around a failed conversion with an installed sidecar.\n\nConvert the final saved Office file to a new PDF in the task workspace. Preserve the source. Quote each input and output path.\n\n```text\n\"\" \"\" convert --input --output \n```\n\nCheck the command result, nonempty output, and `%PDF-` file signature, then report that extent of validation. Unless the user requests deeper checks, do not search for PDF readers, generate images, or install tools to validate a conversion. Spreadsheet PDF pages follow workbook print settings; change those settings only when requested. Conversion may differ from the native application's pagination or fonts.\n\nIf export fails, preserve the usable source and report the export limitation. Do not repeatedly convert an unchanged source or use conversion to investigate a failed preview. For PDF export alone, stop after these checks. Only a request for images or visual inspection calls for `office-render`; load it before that separate operation. Deliver the requested PDF through `present` when available, keeping the file in place.\n\nLibreOffice Kit is disabled in this deployment.\n\n"}],"isError":false,"id":"{{message:8}}"}},"sourceEventSeqs":[20],"surfaceOp":"append"} {"type":"step/end","data":{"turn":1,"step":2}} {"type":"step/start","data":{"turn":1,"step":3}} {"type":"assistant/message","data":{"turn":1,"step":3,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_office_3","name":"skill","arguments":"{\"name\":\"office-recalculate\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:9}}"},"stream":[{"type":"chunk","time":1790753571666,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1790753571666,"index":0,"dt":[],"id":"call_office_3","name":"skill","args":["{\"name\":\"office-recalculate\"}"]},{"type":"chunk","time":1790753571666,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_office_3","name":"skill","arguments":"{\"name\":\"office-recalculate\"}"}}},{"type":"chunk","time":1790753571666,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} From e103591fd5bc2ad8af04e6bca0fc4005b3eeaa0b Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Wed, 14 Oct 2026 13:40:25 +0800 Subject: [PATCH 021/162] fix(subagent): address model selection review feedback --- docs/subsystems/subagent.md | 8 ---- docs/subsystems/subagent.zh.md | 8 ---- docs/tool-catalog.i18n.yaml | 8 ++-- docs/tool-catalog.md | 4 +- docs/tool-catalog.zh.md | 4 +- .../guide.i18n.yaml | 4 +- .../guide.md | 2 + .../guide.zh.md | 2 + .../extensions/tool-cordis/src/api-catalog.ts | 9 +--- .../subagent/tool-subagent/README.i18n.yaml | 4 +- packages/subagent/tool-subagent/README.md | 2 +- packages/subagent/tool-subagent/README.zh.md | 2 +- packages/subagent/tool-subagent/src/index.ts | 9 ++-- .../subagent/tool-subagent/src/list-models.ts | 2 +- .../src/model-selection-settings.ts | 2 +- .../src/model-selection-state.ts | 5 +- .../tool-subagent/src/model-selection.ts | 5 +- .../tests/model-selection-settings.spec.ts | 33 ++++++++----- .../tests/model-selection.spec.ts | 47 +++++++++++++++++-- scripts/gen-tool-catalog.ts | 2 +- .../tool-schemas.expected.json | 4 +- .../tool-schemas.expected.json | 4 +- .../system-prompt.expected.md | 4 +- .../system-prompt.expected.md | 4 +- .../web/ptc-round/system-prompt.expected.md | 4 +- 25 files changed, 112 insertions(+), 70 deletions(-) diff --git a/docs/subsystems/subagent.md b/docs/subsystems/subagent.md index a6910315b3..b54ce57175 100644 --- a/docs/subsystems/subagent.md +++ b/docs/subsystems/subagent.md @@ -406,14 +406,6 @@ Generated from source by `scripts/gen-cordis-catalog.ts` (verified fresh by `pnp Host settings owner that records Session authorization before the first model request. -```ts cordis-catalog -/** - * Read a detached selection preference for the next fresh top-level Session. - * @returns the enabled state and exact allowed routes. - */ -current(): SubagentModelSelectionSettings -``` - Source: [`packages/subagent/tool-subagent/src/model-selection-settings.ts`](../../packages/subagent/tool-subagent/src/model-selection-settings.ts) diff --git a/docs/subsystems/subagent.zh.md b/docs/subsystems/subagent.zh.md index 8c4a34cd1a..0ca87c8503 100644 --- a/docs/subsystems/subagent.zh.md +++ b/docs/subsystems/subagent.zh.md @@ -408,14 +408,6 @@ Generated from source by `scripts/gen-cordis-catalog.ts` (verified fresh by `pnp Host settings owner that records Session authorization before the first model request. -```ts cordis-catalog -/** - * Read a detached selection preference for the next fresh top-level Session. - * @returns the enabled state and exact allowed routes. - */ -current(): SubagentModelSelectionSettings -``` - Source: [`packages/subagent/tool-subagent/src/model-selection-settings.ts`](../../packages/subagent/tool-subagent/src/model-selection-settings.ts) diff --git a/docs/tool-catalog.i18n.yaml b/docs/tool-catalog.i18n.yaml index 72364bdbef..4381738a1e 100644 --- a/docs/tool-catalog.i18n.yaml +++ b/docs/tool-catalog.i18n.yaml @@ -9,8 +9,8 @@ en: 6cc2cd9ced70bb09 zh: 1c0e6cb4ba2ed637 /tool-schema-catalog/tool-package-map: - en: e4b330276e8729aa - zh: 3f781cb32fd16d81 + en: 6db10eb3e60a3015 + zh: 721026cb011ca99e /tool-schema-catalog/deepseek-ai-dsh-plugin-manager: en: e0143916bb634762 zh: e0143916bb634762 @@ -225,8 +225,8 @@ en: 841abe0dd5913ea7 zh: 841abe0dd5913ea7 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/list-subagent-models: - en: 6872bee11fccb6c8 - zh: 230f2d443dc5d9a6 + en: 207228068bc61885 + zh: 36e9344ee1af8869 /tool-schema-catalog/deepseek-ai-dsh-tool-subagent/subagent: en: 5bcfcf9ca4117096 zh: 38cb646632b457b2 diff --git a/docs/tool-catalog.md b/docs/tool-catalog.md index 0ca8251ffd..077253129f 100644 --- a/docs/tool-catalog.md +++ b/docs/tool-catalog.md @@ -37,7 +37,7 @@ This table connects model-visible tool names to the plugin package and service s | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`, `ctx.workflowEngine`, `ctx.subagents`, `ctx.systemPrompt`, `a calling Agent (exec.agent parents every fresh round)` | `tool/call`, `tool/result`, `workflow and child session events during execution` | - | A fixed foreground workflow starts one fresh structured child per round; the model selects only the immutable objective and an optional round cap. | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`, `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave `subagent_fork` route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`, `subagent` | `ctx.tools`, `ctx.subagents`, `ctx.systemPrompt`, `ctx.sessionProjections`, `ctx.llm for model discovery and selected-route validation` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The load-time `toolName` selects the delegation name (default `subagent`); `modelSelection` adds optional route fields. The independent `/list-models` plugin registers discovery. Web and Desktop presets mount discovery, enable spawn route fields, and leave `subagent_fork` route fields disabled. The Host settings service records authorization for each new top-level Session; child Sessions inherit it. Without a recorded policy, discovery and explicit selection permit only the current parent route. Every delegation returns a child id promptly; local children accept follow-ups, while external backends report their single execution automatically. | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named discovery, interruption, and local-message tools over managed subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`, `job_list`, `job_output` | `ctx.tools`, `ctx.jobs`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-job controller: background bash commands and PTY sends are read, listed, and killed through the same three tools. Loading the plugin attaches the controller that arms producers' `ctx.jobs.start()`. | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`, `list_agents`, `send_message`, `spawn_teammate`, `team_task_create`, `team_task_get`, `team_task_list`, `team_task_update`, `wait_agent` | `ctx.tools`, `ctx.systemPrompt`, `ctx.agentTeams`, `an exact live Team member Agent` | `tool/call`, `team/member`, `team/task`, `agent/inbox/spliced`, `user/message on inbox claim`, `tool/result` | - | All nine tools are scoped to implicit Team Leads and durable teammates. The shipped dsh-base bundle keeps the package disabled; the documented Agent Teams profile patch enables it while disabling the legacy continuable-child control names. | @@ -2017,7 +2017,7 @@ The five read-only tools hide provider cursors and authorize every result from t ### `list_subagent_models` -Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. +Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Inspect options here before selecting a child provider and model; use returned ids for delegation. ```json { diff --git a/docs/tool-catalog.zh.md b/docs/tool-catalog.zh.md index 230ebbb0b3..cafe401972 100644 --- a/docs/tool-catalog.zh.md +++ b/docs/tool-catalog.zh.md @@ -41,7 +41,7 @@ | `@deepseek-ai/dsh-tool-ralph` | `ralph` | `ctx.tools`、`ctx.workflowEngine`、`ctx.subagents`、`ctx.systemPrompt`、`a calling Agent (exec.agent parents every fresh round)` | `tool/call`、`tool/result`、`workflow and child session events during execution` | - | 固定的前台工作流会在每个 Round 启动一个全新的结构化子级；模型只能选择不可变目标和可选的 Round 上限。 | | `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.workingDirectory`、`ctx.tools`、`ctx.agents`、`ctx.skills` | `tool/call`、`tool/result`、`user/message replacement catalogs via agent.inject()` | - | - | | `@deepseek-ai/dsh-tool-session-query` | `session_event_read`、`session_event_search`、`session_event_trace`、`session_search`、`session_trace` | `ctx.tools`、`ctx.systemPrompt`、`ctx.sessionQuery`、`a calling Agent for workspace authority` | `tool/call`、`tool/result` | - | 这 5 个只读工具会隐藏提供方游标，并根据不可变的调用 agent 会话为每个结果授权。该包需要选择启用；需要强制截止时间或限制行内输出的组合还会挂载通用超时或 spill 策略。 | -| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 `subagent_fork` 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | +| `@deepseek-ai/dsh-tool-subagent` | `list_subagent_models`、`subagent` | `ctx.tools`、`ctx.subagents`、`ctx.systemPrompt`、`ctx.sessionProjections`、`用于模型发现和所选路由校验的 ctx.llm` | `tool/call`、`tool/result`、`child session events through the chosen provider` | `subagent`、`subagent_fork` | 加载时的 `toolName` 选择委派名称（默认为 `subagent`）；`modelSelection` 增加可选路由字段。独立的 `/list-models` 插件注册发现工具。Web 与 Desktop preset 挂载发现工具，启用 spawn 路由字段，并为 `subagent_fork` 保持路由字段关闭。Host 设置服务为每个新顶层 Session 记录授权；子 Session 继承该授权。没有已记录策略时，发现和显式选择只允许当前父级路由。每次委派都会立即返回子代理 id；本地子代理接受后续消息，外部后端自动汇报单次执行结果。 | | `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`、`list_agents`、`send_message` | `ctx.tools`、`ctx.subagents`、`ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`、`tool/result`、`child session events through ctx.subagents` | - | 这些是用于发现、中断托管 subagent 和向本地子代理发送消息的全局命名工具：绑定提供方的 `tool-subagent` 实例注册不同的委派工具；本包注册一次 `send_message` 和 `interrupt_agent`，另由 `list_agents` 通过单独加载的 `/list-agents` 插件提供，其目录行使用 sessionProjections 和实时 Agent 注册表。 | | `@deepseek-ai/dsh-tool-jobs` | `job_kill`、`job_list`、`job_output` | `ctx.tools`、`ctx.jobs`、`ctx.systemPrompt` | `tool/call`、`tool/result`、`user/message via agent.inject() for background completion notices` | - | 与任务种类无关的后台任务控制器：后台 bash 命令与 PTY 发送 都通过相同的 3 个工具读取、列出和终止。加载该插件会挂接控制器，从而启用生产方的 `ctx.jobs.start()`。 | | `@deepseek-ai/dsh-experimental-tool-agent-team` | `interrupt_agent`、`list_agents`、`send_message`、`spawn_teammate`、`team_task_create`、`team_task_get`、`team_task_list`、`team_task_update`、`wait_agent` | `ctx.tools`、`ctx.systemPrompt`、`ctx.agentTeams`、`an exact live Team member Agent` | `tool/call`、`team/member`、`team/task`、`agent/inbox/spliced`、`user/message on inbox claim`、`tool/result` | - | 这 9 个工具限定于隐式 Team Lead 与持久 teammate 作用域。随产品发布的 dsh-base bundle 默认禁用该包；文档中的 Agent Teams profile patch 会启用它，并禁用旧 continuable child 的同名控制工具。 | @@ -2023,7 +2023,7 @@ lsp 工具将提供方选择和语言服务器子进程置于 ctx.lsp 之后， ### `list_subagent_models` -查询 subagent 可用的模型，不更改当前 Agent。无参数：提供方；`provider`：允许的模型；`provider` + `model`：该模型及其推理强度。将返回的 id 用于委派。 +查询 subagent 可用的模型，不更改当前 Agent。无参数：提供方；`provider`：允许的模型；`provider` + `model`：该模型及其推理强度。选择子级提供方和模型前先在此查询选项，再将返回的 id 用于委派。 ```json { diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml index 5cc66aa405..565e20541f 100644 --- a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml +++ b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.i18n.yaml @@ -9,8 +9,8 @@ en: c54f686a434d8b84 zh: f32b63140d9b2dda /separate-subagent-model-selection-configuration/change: - en: 6773f124a0ff9f6a - zh: 99213e0ba2caff35 + en: 3d948fe459b20e23 + zh: cb3c924d9fde2432 /separate-subagent-model-selection-configuration/migration: en: b83bd0256fb89c92 zh: 5b14cf7cfa9677b3 diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md index 72411e4bc9..d76f6f28e1 100644 --- a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md +++ b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.md @@ -11,6 +11,8 @@ English | [中文](guide.zh.md) `@deepseek-ai/dsh-tool-subagent` replaces `modelSelectionSettings` with `modelSelection`, which controls only optional child route fields. The tool no longer registers `list_subagent_models` or samples user settings. The Host `/model-selection-settings` service records authorization during Agent creation, including Sessions created before delegation tools are mounted. Existing Sessions keep their recorded policy; installing tools later does not resample settings. Shipped Web and Desktop presets include the independent discovery plugin and keep fork route fields disabled. Custom fork tools may enable them. +A Host with model selection enabled and an empty allowed-model list rejects every fresh top-level Session, including those without delegation tools. The settings UI prevents saving that combination; configurations written directly must supply a non-empty list or disable selection. + ## Migration 1. In custom `cordis.yml`, patches, or direct plugin configuration, rename `modelSelectionSettings: true` to `modelSelection: true` on delegation tools that expose route arguments. diff --git a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md index f312728181..336e18d4ee 100644 --- a/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md +++ b/docs/upgrade-guide/v0.2.1-alpha.2/subagent-model-selection-composition/guide.zh.md @@ -11,6 +11,8 @@ description: 自定义 subagent 组合必须分离路由字段、模型发现和 `@deepseek-ai/dsh-tool-subagent` 将 `modelSelectionSettings` 替换为 `modelSelection`，该选项只控制可选子级路由字段。工具不再注册 `list_subagent_models` 或采样用户设置。Host 的 `/model-selection-settings` 服务在 Agent 创建时记录授权，包括在委派工具挂载前创建的 Session。已有 Session 保留已记录策略；之后安装工具不会重新采样设置。随附 Web 与 Desktop preset 包含独立发现插件，并为 fork 保持路由字段关闭。自定义 fork 工具可以启用这些字段。 +Host 启用模型选择但允许模型列表为空时，会拒绝创建所有新的顶层 Session，包括未挂载委派工具的 Session。设置界面不允许保存这种组合；直接写入的配置必须提供非空列表或关闭模型选择。 + ## 迁移 1. 在自定义 `cordis.yml`、patch 或直接插件配置中，将公开路由参数的委派工具的 `modelSelectionSettings: true` 改为 `modelSelection: true`。 diff --git a/packages/extensions/tool-cordis/src/api-catalog.ts b/packages/extensions/tool-cordis/src/api-catalog.ts index 6fb7b5084b..850553cae0 100644 --- a/packages/extensions/tool-cordis/src/api-catalog.ts +++ b/packages/extensions/tool-cordis/src/api-catalog.ts @@ -2958,14 +2958,7 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [ key: 'subagentModelSelection', summary: 'Host settings owner that records Session authorization before the first model request.', description: 'Host settings owner that records Session authorization before the first model request.', - methods: [ - { - signature: 'current(): SubagentModelSelectionSettings', - description: 'Read a detached selection preference for the next fresh top-level Session.', - parameters: [], - returns: 'the enabled state and exact allowed routes.', - }, - ], + methods: [], }, { key: 'subagents', diff --git a/packages/subagent/tool-subagent/README.i18n.yaml b/packages/subagent/tool-subagent/README.i18n.yaml index 621428968a..1290c018fe 100644 --- a/packages/subagent/tool-subagent/README.i18n.yaml +++ b/packages/subagent/tool-subagent/README.i18n.yaml @@ -69,8 +69,8 @@ en: a4ca31b9cbacda6b zh: 91ac3e0784a14697 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/what-the-model-sees: - en: 192670f899eb74f5 - zh: 06eac751268c3a9d + en: 9cf8be0502db31dd + zh: 16eebce205f12364 /deepseek-ai-dsh-tool-subagent/model-experience/model-selection-and-discovery/token-effect: en: 4fa61dc9a52259a2 zh: bc51ac3b1a1cc6ad diff --git a/packages/subagent/tool-subagent/README.md b/packages/subagent/tool-subagent/README.md index 46be412d92..1a4f3b3677 100644 --- a/packages/subagent/tool-subagent/README.md +++ b/packages/subagent/tool-subagent/README.md @@ -152,7 +152,7 @@ Prefix-stable while provider instances and their configuration are unchanged. Ad #### What the model sees -The independent `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. +The independent `list_subagent_models` definition reads the calling Session’s route policy, falling back to the current parent route when no policy is recorded. Calls reject while the optional `ctx.llm` service is unavailable. With a recorded policy, provider model lists intersect the advertised catalog with the Session’s authorized routes. Without one, they resolve only the current parent model, even if the catalog omits it. The discovery tool’s own description instructs the model to query options before selecting a child route; omitting or filtering out discovery removes that guidance. Exact-model queries accept authorized ids absent from the catalog; unauthorized provider/model lookups fail before adapter access. Catalog or exact-model lookup failures and cancellation reject the call. Execution independently enforces the same policy. #### Token effect diff --git a/packages/subagent/tool-subagent/README.zh.md b/packages/subagent/tool-subagent/README.zh.md index d1b9c788e1..5a14b36b6f 100644 --- a/packages/subagent/tool-subagent/README.zh.md +++ b/packages/subagent/tool-subagent/README.zh.md @@ -152,7 +152,7 @@ kind: "package-reference" #### 模型看到什么 -独立的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 +独立的 `list_subagent_models` 定义读取调用 Session 的路由策略，没有已记录策略时使用当前父级路由。可选的 `ctx.llm` 服务不可用时，调用会被拒绝。有已记录策略时，提供方模型列表取公布目录与 Session 授权路由的交集。没有策略时，只解析当前父级模型，即使目录未列出它。发现工具自身的描述要求模型在选择子级路由前查询选项；未挂载或过滤掉发现工具时，这段指导也不会出现。精确模型查询允许已授权但目录未列出的 ID；未授权的提供方／模型查询会在访问适配器前失败。目录或精确模型查询失败以及取消都会使调用被拒绝。执行过程独立强制同一策略。 #### Token 影响 diff --git a/packages/subagent/tool-subagent/src/index.ts b/packages/subagent/tool-subagent/src/index.ts index b7c95a8e1a..03a5fbe514 100644 --- a/packages/subagent/tool-subagent/src/index.ts +++ b/packages/subagent/tool-subagent/src/index.ts @@ -193,8 +193,8 @@ export async function apply(ctx: Context, config: Config): Promise { const wording = providerWording(subagentProvider.inheritsParentContext) const providerRouteDefaults = subagentProvider.agentRouteDefaults const selectionDescription = providerRouteDefaults !== undefined - ? ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend\'s route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' - : ' Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' + ? ' Supply `provider` and `model` together. Omitted values use child defaults, falling back to this backend\'s route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' + : ' Supply `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model\'s default.' const definition = defineTool({ name: toolName, description: wording.description @@ -202,7 +202,10 @@ export async function apply(ctx: Context, config: Config): Promise { + (subagentProvider.prepareContinuable !== undefined ? ' The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes.' : ' The completion notice includes its final answer. This backend does not accept follow-up messages.') - + (modelSelectionEnabled ? selectionDescription : ''), + + (modelSelectionEnabled ? selectionDescription : '') + + (modelSelectionEnabled && subagentProvider.inheritsParentContext + ? ' Changing the route can prevent provider-side reuse of the inherited conversation prefix.' + : ''), parameters: { cwd: { type: 'string', diff --git a/packages/subagent/tool-subagent/src/list-models.ts b/packages/subagent/tool-subagent/src/list-models.ts index 4a81c568f6..f220c85e32 100644 --- a/packages/subagent/tool-subagent/src/list-models.ts +++ b/packages/subagent/tool-subagent/src/list-models.ts @@ -109,7 +109,7 @@ export function apply(ctx: Context): void { description: 'Inspect subagent models without changing the current Agent. No arguments: providers; ' + '`provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. ' - + 'Use returned ids for delegation.', + + 'Inspect options here before selecting a child provider and model; use returned ids for delegation.', parameters: { provider: { type: 'string', diff --git a/packages/subagent/tool-subagent/src/model-selection-settings.ts b/packages/subagent/tool-subagent/src/model-selection-settings.ts index 9e4907556b..a7954c3850 100644 --- a/packages/subagent/tool-subagent/src/model-selection-settings.ts +++ b/packages/subagent/tool-subagent/src/model-selection-settings.ts @@ -72,7 +72,7 @@ export class SubagentModelSelectionConfig extends Service { * Read a detached selection preference for the next fresh top-level Session. * @returns the enabled state and exact allowed routes. */ - current(): SubagentModelSelectionSettings { + private current(): SubagentModelSelectionSettings { const enabled = this.config.enabled.get() const allowedModels = this.config.allowedModels.get() assertAllowedModelRoutes(allowedModels) diff --git a/packages/subagent/tool-subagent/src/model-selection-state.ts b/packages/subagent/tool-subagent/src/model-selection-state.ts index 221ef5a8f6..26516749bd 100644 --- a/packages/subagent/tool-subagent/src/model-selection-state.ts +++ b/packages/subagent/tool-subagent/src/model-selection-state.ts @@ -33,7 +33,10 @@ const modelSelectionPolicySchema: zod.ZodType = zod. model: zod.string().min(1), }).strict()).min(1).nullable() -/** Host-only projection of the durable model-selection policy. */ +/** + * Host-only projection of the durable model-selection policy. + * Each independently mounted consumer registers this shared definition; the registry reference-counts it. + */ export const subagentModelSelectionProjectionDefinition = { key: 'subagentModelSelectionPolicy', stateVersion: 1, diff --git a/packages/subagent/tool-subagent/src/model-selection.ts b/packages/subagent/tool-subagent/src/model-selection.ts index 04f4d33b5f..911901a2ce 100644 --- a/packages/subagent/tool-subagent/src/model-selection.ts +++ b/packages/subagent/tool-subagent/src/model-selection.ts @@ -152,7 +152,10 @@ export function assertAllowedModelSelection( ? provider === parentOptions.provider && model === parentOptions.model : policy.routes.some(route => route.provider === provider && route.model === model) if (allowed) return - throw new Error(`child LLM route "${provider}/${model}" is not allowed for this Session`) + throw new Error(`child LLM route "${provider}/${model}" is not allowed for this Session` + + (policy === undefined + ? '; without a recorded model-selection policy, explicit selection (including reasoning_effort) is limited to the parent route' + : '')) } /** diff --git a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts index 0718087bb6..1b2b11390a 100644 --- a/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection-settings.spec.ts @@ -78,51 +78,62 @@ async function createAgent(ctx: Context, id: string, options: { return handle.agent } +let policySessionCounter = 0 + +/** Observe settings through the policy captured by a fresh top-level Agent. */ +async function capturePolicy(ctx: Context) { + const { agent } = await ctx.agents.create({ sessionId: SessionId(`settings-policy-${++policySessionCounter}`) }) + return subagentModelSelectionPolicy(ctx.sessionProjections, agent.session) +} + describe('SubagentModelSelectionConfig', () => { it('uses the composed default without a settings provider', async () => { const ctx = ownContext(new Context()) - await mountAgentLoopTestDependencies(ctx) + await mountAgentLoopTestDependencies(ctx, { workingDirectory: true }) + await ctx.plugin(AgentLoop, { agents: [] }) await ctx.plugin(SubagentModelSelectionConfig, { enabled: true, allowedModels: ALLOWED_MODELS }) - expect(ctx.subagentModelSelection.current()).toEqual({ enabled: true, allowedModels: ALLOWED_MODELS }) + expect(await capturePolicy(ctx)).toEqual(ALLOWED_MODELS) await ctx.fiber.dispose() }) it('defaults off and follows the validated user layer', async () => { const ctx = ownContext(new Context()) - await mountAgentLoopTestDependencies(ctx) + await mountAgentLoopTestDependencies(ctx, { workingDirectory: true }) + await ctx.plugin(AgentLoop, { agents: [] }) selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) - expect(ctx.subagentModelSelection.current()).toEqual({ enabled: false, allowedModels: [] }) + expect(await capturePolicy(ctx)).toBeUndefined() await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: ALLOWED_MODELS, }) - expect(ctx.subagentModelSelection.current()).toEqual({ enabled: true, allowedModels: ALLOWED_MODELS }) + expect(await capturePolicy(ctx)).toEqual(ALLOWED_MODELS) await ctx.fiber.dispose() }) it('rejects duplicate routes, enabled empty settings, and an empty durable policy', async () => { const ctx = ownContext(new Context()) - await mountAgentLoopTestDependencies(ctx) + await mountAgentLoopTestDependencies(ctx, { workingDirectory: true }) + await ctx.plugin(AgentLoop, { agents: [] }) selectionConfigs.set(ctx, await liveConfig(ctx, SubagentModelSelectionConfig)) ctx.sessionProjections.register(subagentModelSelectionProjectionDefinition) - // A stored policy the service cannot apply fails when it is read, not when the profile row is written. + // File-written invalid settings reject Agent creation even without delegation tools. await selectionConfigs.get(ctx)!.update({ allowedModels: [...ALLOWED_MODELS, ...ALLOWED_MODELS] }) - expect(() => ctx.subagentModelSelection.current()).toThrow('repeats route "alpha/fast-model"') + await expect(capturePolicy(ctx)).rejects.toThrow('repeats route "alpha/fast-model"') await selectionConfigs.get(ctx)!.update({ enabled: true, allowedModels: [] }) - expect(() => ctx.subagentModelSelection.current()).toThrow('enabled subagent model selection requires at least one allowed model') + await expect(capturePolicy(ctx)).rejects.toThrow('enabled subagent model selection requires at least one allowed model') await selectionConfigs.get(ctx)!.update({ enabled: false, allowedModels: ALLOWED_MODELS, }) - expect(ctx.subagentModelSelection.current()).toEqual({ enabled: false, allowedModels: ALLOWED_MODELS }) + expect(await capturePolicy(ctx)).toBeUndefined() await selectionConfigs.get(ctx)!.update({ allowedModels: [] }) - expect(ctx.subagentModelSelection.current()).toEqual({ enabled: false, allowedModels: [] }) + expect(await capturePolicy(ctx)).toBeUndefined() const invalid = Session.create(SessionId('empty-policy')) invalid.append('subagent/model-selection-policy', { allowedModels: [] }) diff --git a/packages/subagent/tool-subagent/tests/model-selection.spec.ts b/packages/subagent/tool-subagent/tests/model-selection.spec.ts index 183f1f52f8..6a437cea37 100644 --- a/packages/subagent/tool-subagent/tests/model-selection.spec.ts +++ b/packages/subagent/tool-subagent/tests/model-selection.spec.ts @@ -12,6 +12,7 @@ import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection' import { MockAdapter } from '../../../core/agent-loop/tests/mock-adapter.ts' import * as mock from './scripted-provider.ts' import * as tool from '../src/index.ts' +import * as ListModels from '../src/list-models.ts' import { assertAllowedModelRoutes, assertAllowedModelSelection, @@ -121,7 +122,8 @@ describe('dsh-tool-subagent model selection', () => { 'provider', 'reasoning_effort', ]) - expect(schema.description).toContain('list_subagent_models') + expect(schema.description).not.toContain('list_subagent_models') + expect(schema.description).not.toContain('Changing the route can prevent provider-side reuse') expect(ctx.tools.get('list_subagent_models', agent)).toBeDefined() expect(schema.description).not.toContain('alpha') @@ -159,6 +161,43 @@ describe('dsh-tool-subagent model selection', () => { } }) + it('shows discovery guidance only while its tool is visible, regardless of mount order', async () => { + const ctx = await setup({ provider: 'mock', modelSelection: true }) + const agent = modelSelectionSetupAgent(ctx) + const descriptions = () => ctx.tools.schemas(agent).map(schema => schema.description).join('\n') + const guidance = 'Inspect options here before selecting a child provider and model' + expect(descriptions()).not.toContain(guidance) + expect(descriptions()).not.toContain('list_subagent_models') + + const discovery = await ctx.plugin(ListModels) + expect(descriptions()).toContain(guidance) + const restore = agent.ctx.tools.restrict({ deny: ['list_subagent_models'] }) + expect(descriptions()).not.toContain(guidance) + restore() + expect(descriptions()).toContain(guidance) + await discovery.dispose() + expect(descriptions()).not.toContain(guidance) + }) + + it.each(['configured', 'provider'] as const)('explains effort-only denial on a %s route without Session authorization', async (source) => { + const defaults = { provider: 'alpha', model: 'configured-model' } + const start = vi.fn() + const ctx = await setup({ + provider: 'mock', modelSelection: true, + parentAgentOptions: { provider: 'alpha', model: 'parent-model' }, + ...source === 'configured' ? { agentOptions: defaults } : {}, + }, { + onStart: start, + ...source === 'provider' ? { agentRouteDefaults: defaults } : {}, + }) + const result = await callSubagent(ctx, { + description: 'adjust effort', prompt: 'work', reasoning_effort: 'low', + }) + expect(result.isError).toBe(true) + expect(text(result)).toContain('without a recorded model-selection policy, explicit selection (including reasoning_effort) is limited to the parent route') + expect(start).not.toHaveBeenCalled() + }) + it('keeps settings-owned route fields visible but rejects other routes without a Session policy', async () => { const ctx = await setup({ provider: 'mock', @@ -170,7 +209,8 @@ describe('dsh-tool-subagent model selection', () => { const schema = ctx.tools.schemas(agent).find(entry => entry.name === 'subagent')! const props = (schema.parameters as { properties?: Record }).properties ?? {} expect(Object.keys(props).sort()).toEqual(['cwd', 'description', 'model', 'prompt', 'provider', 'reasoning_effort']) - expect(schema.description).toContain('list_subagent_models') + expect(schema.description).not.toContain('list_subagent_models') + expect(schema.description).not.toContain('Changing the route can prevent provider-side reuse') expect(ctx.tools.get('list_subagent_models', agent)).toBeDefined() const result = await callSubagent(ctx, { @@ -505,7 +545,8 @@ describe('dsh-tool-subagent model selection', () => { const agent = modelSelectionSetupAgent(ctx) const schema = ctx.tools.schemas(agent).find(entry => entry.name === 'subagent')! expect(schema.description).toContain('inherits this conversation') - expect(schema.description).toContain('list_subagent_models') + expect(schema.description).not.toContain('list_subagent_models') + expect(schema.description).toContain('Changing the route can prevent provider-side reuse of the inherited conversation prefix.') const allowed = await callSubagent(ctx, { description: 'fork work', prompt: 'do it', provider: 'alpha', model: 'fast-model', reasoning_effort: 'low', diff --git a/scripts/gen-tool-catalog.ts b/scripts/gen-tool-catalog.ts index 3f3288a6b8..df595823fe 100644 --- a/scripts/gen-tool-catalog.ts +++ b/scripts/gen-tool-catalog.ts @@ -511,7 +511,7 @@ const TOOL_PACKAGES: ToolPackage[] = [ list_subagent_models: 'packages/subagent/tool-subagent/src/list-models.ts', subagent: 'packages/subagent/tool-subagent/src/index.ts', }, - requires: ['ctx.tools', 'ctx.subagents', 'ctx.systemPrompt', 'ctx.llm for model discovery and selected-route validation'], + requires: ['ctx.tools', 'ctx.subagents', 'ctx.systemPrompt', 'ctx.sessionProjections', 'ctx.llm for model discovery and selected-route validation'], writes: ['tool/call', 'tool/result', 'child session events through the chosen provider'], shippedNames: ['subagent', 'subagent_fork'], async mount(ctx) { diff --git a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json index 85e10ae4f2..28da314b29 100644 --- a/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json +++ b/snapshots/sdk/subagent-dsh-sdk-dynamic-route/tool-schemas.expected.json @@ -262,7 +262,7 @@ }, { "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Inspect options here before selecting a child provider and model; use returned ids for delegation.", "parameters": { "type": "object", "properties": { @@ -356,7 +356,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The completion notice includes its final answer. This backend does not accept follow-up messages. Supply `provider` and `model` together. Omitted values use child defaults, falling back to this backend's route defaults. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { diff --git a/snapshots/web/fresh-round-trip/tool-schemas.expected.json b/snapshots/web/fresh-round-trip/tool-schemas.expected.json index 34cc3505b3..d97f06bdb1 100644 --- a/snapshots/web/fresh-round-trip/tool-schemas.expected.json +++ b/snapshots/web/fresh-round-trip/tool-schemas.expected.json @@ -325,7 +325,7 @@ }, { "name": "list_subagent_models", - "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation.", + "description": "Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Inspect options here before selecting a child provider and model; use returned ids for delegation.", "parameters": { "type": "object", "properties": { @@ -719,7 +719,7 @@ }, { "name": "subagent", - "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", + "description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Supply `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default.", "parameters": { "type": "object", "properties": { diff --git a/snapshots/web/official-subagent-bundle/system-prompt.expected.md b/snapshots/web/official-subagent-bundle/system-prompt.expected.md index bf81e7f52a..778c016948 100644 --- a/snapshots/web/official-subagent-bundle/system-prompt.expected.md +++ b/snapshots/web/official-subagent-bundle/system-prompt.expected.md @@ -158,7 +158,7 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Inspect options here before selecting a child provider and model; use returned ids for delegation. */ list_subagent_models: { /** Registered LLM provider id. */ provider?: string; @@ -288,7 +288,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Supply `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; diff --git a/snapshots/web/optional-string-editor/system-prompt.expected.md b/snapshots/web/optional-string-editor/system-prompt.expected.md index d1d3f1d372..1670257a1e 100644 --- a/snapshots/web/optional-string-editor/system-prompt.expected.md +++ b/snapshots/web/optional-string-editor/system-prompt.expected.md @@ -158,7 +158,7 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Inspect options here before selecting a child provider and model; use returned ids for delegation. */ list_subagent_models: { /** Registered LLM provider id. */ provider?: string; @@ -305,7 +305,7 @@ interface ToolArgsMap { /** Optional parameter of `view` command when `path` points to a file. If omitted or null, the full file is shown. If provided, the file will be shown in the indicated line number range, e.g. [11, 12] will show lines 11 and 12. Indexing at 1 to start. Setting `[start_line, -1]` shows all lines from `start_line` to the end of the file. */ view_range?: number[] | null; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Supply `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; diff --git a/snapshots/web/ptc-round/system-prompt.expected.md b/snapshots/web/ptc-round/system-prompt.expected.md index 8818cb5cce..28e71009d3 100644 --- a/snapshots/web/ptc-round/system-prompt.expected.md +++ b/snapshots/web/ptc-round/system-prompt.expected.md @@ -158,7 +158,7 @@ interface ToolArgsMap { /** children (default) lists direct children, which accept send_message in any status. descendants lists the whole tree below you with each entry's parent session id and depth; entries deeper than 1 accept only interrupt_agent. */ scope?: "children" | "descendants"; } & Record; - /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Use returned ids for delegation. */ + /** Inspect subagent models without changing the current Agent. No arguments: providers; `provider`: allowed models; `provider` + `model`: that model and its reasoning efforts. Inspect options here before selecting a child provider and model; use returned ids for delegation. */ list_subagent_models: { /** Registered LLM provider id. */ provider?: string; @@ -288,7 +288,7 @@ interface ToolArgsMap { /** The exact skill name from the available skills list. */ name: string; } & Record; - /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Inspect options with `list_subagent_models` before setting `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ + /** Delegate a self-contained task to a subagent (a separate agent that works in its own context) to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent returns its result, not its intermediate steps. This tool starts an independently managed subagent and immediately returns its id. The runtime notifies you when it finishes. The child reports results with `send_message`; use `send_message` to steer it while running or continue its conversation after it finishes. Supply `provider` and `model` together. Omitted values use child defaults, falling back to compatible parent values. When the effective route changes, an omitted `reasoning_effort` uses the model's default. */ subagent: { /** Initial child working directory. Relative paths use your current directory; omitted inherits it. Later directory changes in either agent are independent. */ cwd?: string; From c5d227001e548422a03e8f1030b80888a3ae337a Mon Sep 17 00:00:00 2001 From: Dudu-0223 Date: Wed, 14 Oct 2026 13:46:35 +0800 Subject: [PATCH 022/162] test(subagent): refresh fork runtime context ordering --- snapshots/web/subagent-fork-schedule/session.1.v4.jsonl | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/snapshots/web/subagent-fork-schedule/session.1.v4.jsonl b/snapshots/web/subagent-fork-schedule/session.1.v4.jsonl index 080926a704..b09cd32916 100644 --- a/snapshots/web/subagent-fork-schedule/session.1.v4.jsonl +++ b/snapshots/web/subagent-fork-schedule/session.1.v4.jsonl @@ -25,7 +25,7 @@ {"type":"agent/inbox/spliced","data":{"target":"next-turn","start":0,"removedCount":1,"inserted":[]}} {"type":"step/start","data":{"turn":2,"step":1}} {"type":"user/message","data":{"content":[{"type":"text","text":"Call schedule_create, schedule_list, schedule_update, and schedule_delete once each, then reply CHILD_DONE."}],"source":{"kind":"user"},"role":"user","id":"{{message:6}}"},"surfaceOp":"append"} -{"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: workspace-write. Any available operation enforced by the DSH file sandbox may modify files under the session workspace: \"{{cwd}}\". Some platform temporary areas may also be writable.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).\n\nDelegation depth for `subagent_fork`: current 1, maximum 1, remaining levels 0. Do not call `subagent_fork` at this depth; carry out the work directly.\n\nYou are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it.\n\nDelegation depth for `subagent`: current 1, maximum 1, remaining levels 0. Do not call `subagent` at this depth; carry out the work directly.\n\nCurrent DeepSeek Harness implementation checkout: \"{{sourceRoot}}\".\n\nCurrent DeepSeek Harness Web GUI URL: \"{{webUrl}}\"."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: workspace-write. Any available operation enforced by the DSH file sandbox may modify files under the session workspace: \"{{cwd}}\". Some platform temporary areas may also be writable."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."},{"name":"tool:subagent_fork:depth","text":"Delegation depth for `subagent_fork`: current 1, maximum 1, remaining levels 0. Do not call `subagent_fork` at this depth; carry out the work directly."},{"name":"subagent:delegation","text":"You are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."},{"name":"tool:subagent:depth","text":"Delegation depth for `subagent`: current 1, maximum 1, remaining levels 0. Do not call `subagent` at this depth; carry out the work directly."},{"name":"harness:source","text":"Current DeepSeek Harness implementation checkout: \"{{sourceRoot}}\"."},{"name":"app:web-surface","text":"Current DeepSeek Harness Web GUI URL: \"{{webUrl}}\"."}]},"role":"user","id":"{{message:7}}"},"surfaceOp":"append"} +{"type":"user/message","data":{"content":[{"type":"text","text":"Current runtime context. This snapshot supersedes earlier runtime-context snapshots.\n\nCurrent working directory: \"{{cwd}}\".\n\nCurrent DSH file policy: workspace-write. Any available operation enforced by the DSH file sandbox may modify files under the session workspace: \"{{cwd}}\". Some platform temporary areas may also be writable.\n\nApproval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).\n\nDelegation depth for `subagent`: current 1, maximum 1, remaining levels 0. Do not call `subagent` at this depth; carry out the work directly.\n\nDelegation depth for `subagent_fork`: current 1, maximum 1, remaining levels 0. Do not call `subagent_fork` at this depth; carry out the work directly.\n\nYou are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it.\n\nCurrent DeepSeek Harness implementation checkout: \"{{sourceRoot}}\".\n\nCurrent DeepSeek Harness Web GUI URL: \"{{webUrl}}\"."}],"source":{"kind":"runtime-context","form":"snapshot","sections":[{"name":"working-directory:current","text":"Current working directory: \"{{cwd}}\"."},{"name":"sandbox:policy","text":"Current DSH file policy: workspace-write. Any available operation enforced by the DSH file sandbox may modify files under the session workspace: \"{{cwd}}\". Some platform temporary areas may also be writable."},{"name":"approval:policy","text":"Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`)."},{"name":"tool:subagent:depth","text":"Delegation depth for `subagent`: current 1, maximum 1, remaining levels 0. Do not call `subagent` at this depth; carry out the work directly."},{"name":"tool:subagent_fork:depth","text":"Delegation depth for `subagent_fork`: current 1, maximum 1, remaining levels 0. Do not call `subagent_fork` at this depth; carry out the work directly."},{"name":"subagent:delegation","text":"You are a delegated subagent: your permission scope was fixed when you were started and cannot be widened from inside this session — operations that require approval are rejected automatically. When the task needs access beyond that scope, do not retry the denied operation; state the limitation in your reply so the delegating agent can handle it."},{"name":"harness:source","text":"Current DeepSeek Harness implementation checkout: \"{{sourceRoot}}\"."},{"name":"app:web-surface","text":"Current DeepSeek Harness Web GUI URL: \"{{webUrl}}\"."}]},"role":"user","id":"{{message:7}}"},"surfaceOp":"append"} {"type":"request/header","data":{"header":{"config":{"provider":"deepseek-official","model":"deepseek-v4-flash"},"tools":"{{tools}}"},"reason":"resume"}} {"type":"assistant/message","data":{"turn":2,"step":1,"message":{"role":"assistant","content":[{"type":"tool-call","id":"call_schedule_create","name":"schedule_create","arguments":"{\"title\":\"Child reminder\",\"prompt\":\"Check this\",\"after_seconds\":60}"},{"type":"tool-call","id":"call_schedule_list","name":"schedule_list","arguments":"{}"},{"type":"tool-call","id":"call_schedule_update","name":"schedule_update","arguments":"{\"id\":\"missing-reminder\",\"title\":\"Updated\"}"},{"type":"tool-call","id":"call_schedule_delete","name":"schedule_delete","arguments":"{\"id\":\"missing-reminder\"}"}],"source":{"kind":"model","provider":"deepseek-official","model":"deepseek-v4-flash"},"id":"{{message:8}}"},"stream":[{"type":"chunk","time":1791527273612,"chunk":{"type":"block-start","index":0,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791527273612,"index":0,"dt":[],"id":"call_schedule_create","name":"schedule_create","args":["{\"title\":\"Child reminder\",\"prompt\":\"Check this\",\"after_seconds\":60}"]},{"type":"chunk","time":1791527273612,"chunk":{"type":"block-end","index":0,"block":{"type":"tool-call","id":"call_schedule_create","name":"schedule_create","arguments":"{\"title\":\"Child reminder\",\"prompt\":\"Check this\",\"after_seconds\":60}"}}},{"type":"chunk","time":1791527273612,"chunk":{"type":"block-start","index":1,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791527273612,"index":1,"dt":[],"id":"call_schedule_list","name":"schedule_list","args":["{}"]},{"type":"chunk","time":1791527273612,"chunk":{"type":"block-end","index":1,"block":{"type":"tool-call","id":"call_schedule_list","name":"schedule_list","arguments":"{}"}}},{"type":"chunk","time":1791527273612,"chunk":{"type":"block-start","index":2,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791527273612,"index":2,"dt":[],"id":"call_schedule_update","name":"schedule_update","args":["{\"id\":\"missing-reminder\",\"title\":\"Updated\"}"]},{"type":"chunk","time":1791527273612,"chunk":{"type":"block-end","index":2,"block":{"type":"tool-call","id":"call_schedule_update","name":"schedule_update","arguments":"{\"id\":\"missing-reminder\",\"title\":\"Updated\"}"}}},{"type":"chunk","time":1791527273612,"chunk":{"type":"block-start","index":3,"blockType":"tool-call"}},{"type":"tool-call-chunks","time0":1791527273612,"index":3,"dt":[],"id":"call_schedule_delete","name":"schedule_delete","args":["{\"id\":\"missing-reminder\"}"]},{"type":"chunk","time":1791527273613,"chunk":{"type":"block-end","index":3,"block":{"type":"tool-call","id":"call_schedule_delete","name":"schedule_delete","arguments":"{\"id\":\"missing-reminder\"}"}}},{"type":"chunk","time":1791527273613,"chunk":{"type":"finish","reason":{"kind":"tool-calls"}}}]},"surfaceOp":"append"} {"type":"tool/call","data":{"turn":2,"step":1,"callId":"call_schedule_create","name":"schedule_create","arguments":"{\"title\":\"Child reminder\",\"prompt\":\"Check this\",\"after_seconds\":60}"}} From c0bf2ac81dc9238257ea602e82913bd11057227a Mon Sep 17 00:00:00 2001 From: Shigma Date: Wed, 14 Oct 2026 14:28:16 +0800 Subject: [PATCH 023/162] docs(cordis): plan the staged rc.11 vendor upgrade --- vendor/README.md | 2 + vendor/cordis-rc11-upgrade-plan.md | 93 ++++++++++++++++++++++++++++++ 2 files changed, 95 insertions(+) create mode 100644 vendor/cordis-rc11-upgrade-plan.md diff --git a/vendor/README.md b/vendor/README.md index 1765056c37..654276201d 100644 --- a/vendor/README.md +++ b/vendor/README.md @@ -6,6 +6,8 @@ All vendored packages use the **`@deepseek-ai` scope** (`cordis` → `@deepseek- This file covers the manifest, the local-modification log, and the procedure for **updating** an existing vendored package. To **add a new** one, see the cookbook guide: [docs/cookbook/adding-a-vendored-package.md](../docs/cookbook/adding-a-vendored-package.md). +The proposed [Cordis rc.11 upgrade plan](cordis-rc11-upgrade-plan.md) records the package scope, PR dependencies and intended disposition of the local modifications. + ## Manifest Upstream workspace: `cordis-workspace` (local checkout: `~/repos/cordis-workspace`). diff --git a/vendor/cordis-rc11-upgrade-plan.md b/vendor/cordis-rc11-upgrade-plan.md new file mode 100644 index 0000000000..ac53067459 --- /dev/null +++ b/vendor/cordis-rc11-upgrade-plan.md @@ -0,0 +1,93 @@ +# Cordis rc.11 vendor upgrade plan + +Status: proposed + +## Problem + +The [vendor manifest](README.md) records Cordis core 4.0.0-rc.7, Loader 1.0.0-rc.5, Include 1.0.4, HMR 1.0.15 and Timer 1.1.2, with 24 local-modification entries. [Cordis v4.0.0-rc.11](https://github.com/cordiverse/cordis/releases/tag/v4.0.0-rc.11) incorporates some fixes and replaces other mechanisms. Updating every package and retiring every patch in one change would combine configuration, persistence, lifecycle and module-replacement decisions. + +DSH also has consumers that upstream Cordis does not own. Volatile configuration fields are stable references updated without restarting plugins; the settings UI reads their live values. A profile selects ordered plugin bundles and configuration overrides. [DSH HMR](../packages/boot/hmr/README.md) coordinates module replacement and profile changes through `ctx.hmr`. An entry's module namespace is its raw import result, retained before selecting `default` or `apply` so HMR can identify the module already loaded. These consumers must remain usable at every merge point. + +## Proposal + +Keep source vendoring and stage the upgrade as independently reviewable PRs, each covering a complete concern with its producers and consumers. Upgrade Core with the minimum cross-package compatibility changes, then complete Loader, Include and vendor HMR separately. Retire an existing patch when its replacement is adopted and verified; retain independently applicable patches until their behavior has a replacement. PR 0 records this plan only. It does not approve the open runtime decisions or implement the upgrade. + +### Package scope + +| Package | Target from the rc.11 tag | Treatment | +|---|---|---| +| `vendor/cordis` | 4.0.0-rc.11 | Upgrade with the Core compatibility slice. | +| `vendor/loader` | 1.0.0-rc.8 | Import required slices first; complete the package after resolution and change reporting land. | +| `vendor/include` | 1.1.1 | Separate patch composition from journal adoption. | +| `vendor/hmr` | 1.1.1 | Upgrade independently of the Include journal. | +| `vendor/timer` | 1.1.4 | Upgrade independently of Core. | +| `vendor/group` | Keep the existing pin | Its entry source matches the tag after rescoping. The Group implementation lives in Loader and still requires migration. | +| `vendor/logger-console` | Keep the existing pin | All three source files match the tag after normalizing scope, relative import extensions and comments. Core's logger changes remain part of the Core upgrade. | +| `vendor/cosmokit`, `vendor/schemastery` | Keep the existing pins | Retain the volatile extensions; their upstream upgrades are outside this plan. | + +The comparison above checks the current vendored source against the release tag; it does not assume equal version numbers imply equal code. DSH package versions, exports and build settings remain governed by the repository. A partially synchronized Loader, Include or HMR keeps its old manifest baseline and records the imported upstream slice in the local-modification log. Advance the package baseline only when the complete package has been reconciled, with remaining differences recorded explicitly. + +### PR dependencies + +The identifiers below are plan labels, not GitHub issue numbers. Preconditions are merge requirements. Independent PRs can be developed and reviewed concurrently; a dependent PR can start against the proposed changes before its predecessors merge, but must validate the actual combined result before merging. + +`Vn` below identifies item `n` in the [current local-modification log](README.md#local-modifications), not a PR. **Retire** means the upstream implementation replaces the local difference; **partial** means some behavior still needs a local adaptation; **retain** means the requirement remains; **prepare** does not retire a patch. These are proposed outcomes, confirmed only by the owning PR's tests. Track each item by its subject if later manifest edits renumber the log. + +| PR | Scope | Prerequisite PRs | Existing vendor modifications | Required evidence | +|---|---|---|---|---| +| 0 | This plan, package comparison and open decisions. | None | None retired; records the proposed disposition. | Documentation checks; no runtime changes. | +| 1 | Define access to raw config, candidate resolution and the active instance's effective config, and migrate settings, Config Editor, volatile updates and HMR readers together. Initially delegate to the old framework fields. | 0 | Prepare V15 (lazy/raw config) and V22 (volatile); retire neither yet. | Preserve existing objects, validation, settings and volatile behavior. Candidate validation does not replace the active instance's references; introduce no dependency from vendor code to DSH packages. | +| 2 | Add upstream-compatible `hmr.watch()` to both HMR providers with compatible typing, migrate the existing Include to explicit registration, and remove the equivalent Include scans. Use DSH HMR's existing coordination queue. | 0 | Adapt V9's config-path matching to explicit watches; retain its readiness and native-path requirements. | Callback ordering, disposal, explicit paths and `root: []` under the old Core; one Include refresh per change under either provider. Preserve profile watching and both module-replacement algorithms. | +| 3 | Upgrade Timer to 1.1.4, retaining DSH locale resources and build metadata. | 0 | Retain V24 (display metadata); no logged runtime patch retired. | Concurrent interval reads, throttle cancellation and void return types; check callers using the old Core. | +| 4a | Upgrade Core to rc.11 with the minimum Loader, Include and HMR compatibility slice described below. | 1 | Retire V12 (activation observer), V21 (logger disposer) and, after canonical-fiber audits, V20 (Entry identity workaround). Partial V6 (lifecycle) and V15 (config model). Retain/adapt V22 and V23 (module namespace). | Config resolution, volatile identity, service isolation, dependent-before-provider teardown, retained effect regressions and application startup. | +| 4b | Import Loader's project-relative bare-specifier resolver, runtime API detection and fallback when Node internals are absent. | 0 | Retire V19: upstream now detects the internal loader by its available APIs. | Resolve in the config project's package scope, retain subpaths, handle read-only helper creation and test Node 24's older v1 API. | +| 4c | Complete Loader's `EntryChange` / `commit(change)` reporting and remaining changes together with upstream Include's pure patch implementation and journal helper module. Adapt the old writer and in-memory trees; keep `applyEntryPatches` and `entryListSchema` exports. | 4a, 4b | Retire V18 (`disabled` expressions). Partial V11: retire the duplicate patch algorithm and insert-index fix; retain public exports. Retain/adapt V22, including the schema receiver fix, and V23. | Create, move, remove and self-update reports; no false self-disposal writes; later patches address inserted rows; cloning and config dumps agree with mounting. The old Include writer remains active. | +| 4e | Complete vendor HMR's upgrade to 1.1.1, preserving required local path, readiness and explicit-watch behavior. | 2, 3, 4a | Partial V9: adopt upstream ordering and path handling where equivalent; retain explicit readiness and realpath gaps. Retain V1 (no YAML locale imports). | Three-stage module replacement, nested entries, disabled/removed entries, explicit watches and the existing local path cases. | +| 4f | Upgrade Include to its journal implementation, retaining the explicit watch registration already adopted in PR 2. | 2, 4c | Retire V8 (old refresh implementation) and V13 (`writeTask` type). Partial V14: replace the write queue; decide unsaved-error reporting. Retain the remaining V11 exports and V15 child-expression marker. | File/runtime conflict handling, patch-owned writes, shutdown results, nested expressions and one coordinated refresh under either HMR provider. | + +Shared integration differences remain in every affected sync: V2 (manifests), V3 (TypeScript projects), V4/V10 (imports and Node-compatible modules), V5 (build overrides), V7 (API documentation), V16 (published Core source) and V17 (scope). Untouched packages retain them without edits. V6 only loses the stale-epoch guard already supplied upstream; its remaining ownership and cleanup guarantees stay. V15 replaces the old raw/effective-config and event implementation, but the Include child-expression behavior remains local. V22 is not upstreamed by this release, including its schema-bound `simplify` call. + +After PR 0, PRs 1, 2, 3 and 4b have no implementation dependency on one another. Core depends on configuration preparation; it does not depend on Timer. Loader completion and vendor HMR can proceed independently after Core. Include journal adoption depends on Loader and the complete watch migration, not on the new vendor HMR replacement algorithm; PRs 4e and 4f can merge in either order. No single GitHub stack should impose a dependency between unrelated branches. Shared files such as the vendor manifest may still require merge-conflict resolution. + +### Minimum Core compatibility slice + +PR 4a must migrate every caller of the changed Core APIs in the same PR, including callers outside vendor. This is a complete Core upgrade with local patches, paired with selected Loader changes; it is not a claim that Loader already equals 1.0.0-rc.8. + +- Adopt upstream raw `Fiber.config`, `internal/config(fiber, next)`, `internal/commit` and the new `internal/update` arguments and awaited result. Update the config accessors, provider hooks, boot observers and old Include's update listener together. The temporary commit listener writes through the old tree writer until PR 4c. +- Carry the required Loader entry/context awaiting chain, activation-failure observer and isolation changes with Core's `Impl.ctx`, `Impl.isolate`, `committed` and dependent tracking. Update direct implementation readers such as the preset registry. Include Group behavior owned by `vendor/loader/src/config/group.ts`; do not confuse it with the unchanged `vendor/group` entry package. Before retiring V20, ensure volatile updates and every other direct lifecycle-field access use the canonical `fiber.ctx.fiber`. +- Keep Include's nested-entry expressions literal until the owning child resolves them. Keep the namespace field needed by DSH HMR and preserve active volatile references through the agreed config mechanism. Update raw-config access in both HMR implementations without replacing the DSH module-replacement policy. +- Reapply the still-needed effect ownership, reentrant disposal and observer-containment guarantees to the new Core lifecycle. Remove patches already supplied by the imported Core source. Verify the combination with provider/dependent teardown before treating the compatibility slice as complete. + +Module resolution, structured tree-change reporting, journal reconciliation and the new vendor HMR algorithm are not prerequisites for these Core interfaces. Their PRs remain separate unless executable compatibility tests expose an additional dependency; amend this plan before broadening a PR silently. + +### Temporary compatibility code + +PR 4c changes tree mutation producers to `commit(change)` while the old Include remains installed. That Include can implement `commit(change)` by invoking its existing `write()` operation; it deliberately ignores the detailed change report until the journal replaces the writer. In-memory trees, including `PresetTree`, implement a no-op commit. This avoids maintaining two mutation-reporting protocols in Loader. PR 4f deletes the Include writer adapter and old queue. + +PR 4c imports `patch.ts` together with its `journal.ts` helpers and their Loader `EntryChange` dependency, even though Include does not start runtime journaling yet. The public patch export may remain an alias; it must not retain a second algorithm. Every temporary divergence has an owning introduction PR and a named removal PR. No runtime version detection or second Core event protocol is planned. + +PR 2 moves Include refresh ownership in one step: both HMR providers expose `watch()`, the existing Include registers its callback, and both providers stop their equivalent Include scans. The upstream HMR source no longer performs that scan, so importing it before migrating the Include consumer would lose config refresh. Moving the complete watch concern first avoids a temporary scan adapter and makes the later module-replacement and persistence decisions independent. + +### Decisions owned by individual PRs + +| Decision | Owning PR | Proposal and unresolved choice | +|---|---|---| +| Active effective config | 1, 4a | Store or expose the object actually supplied to the current instance. Resolving a candidate through `internal/config` must not overwrite it. Decide the publication point and invalidation rules before switching Core; a WeakMap alone does not establish those rules. | +| Persistence outcome | 4f | Prefer the journal's retry and reconciliation mechanism. Decide how explicit callers observe modifications still unsaved at shutdown. Do not require a historical failure to remain fatal after a successful retry merely because the old test expects it. | +| Nested config interpolation | 4a, 4f | Preserve child-owned expression evaluation. Keep the existing Include marker initially; any refinement that permits interpolation of Include's own fields while excluding child lists is a separate behavior change. | +| HMR replacement failure | 4e | Adopt upstream behavior for vendor HMR. Preserve DSH HMR's current source-reload rollback policy in this series; replacing that policy requires its own proposal and tests. | +| Effect cleanup guarantees | 4a | Retain guarantees not supplied by rc.11. Further upstreaming or retirement is separate work, not a prerequisite for reaching the release baseline. | + +## Acceptance criteria + +- Every implementation PR leaves supported applications usable and passes checks appropriate to its changes. A later PR may not be required to repair an earlier PR's startup, types, persistence or teardown. +- Each source-vendor update follows the [sync procedure](README.md#sync-procedure), including `pnpm run test` and `pnpm run build`, and updates its own manifest, local-modification log, affected documentation and generated references. The [pre-push workflow](../.agents/skills/dsh-pre-push-checks/SKILL.md) selects additional focused checks. Documentation freshness is not deferred to a final cleanup PR. +- Every changed behavior has an explicit decision and focused regression coverage. Tests for an intentionally superseded behavior change with its implementation and explain the new expectation. +- Once all implementation PRs are complete, Core, Loader, Include, vendor HMR and Timer use their selected rc.11-tag sources with enumerated remaining patches. Group, logger-console, Cosmokit and Schemastery keep their existing pins. All temporary writer and config adapters have been removed or reduced to the agreed permanent accessors. +- Final integration verification exercises settings and volatile updates, profile recomposition, both HMR providers, Include persistence and representative built application launches. This is an acceptance checkpoint, not an otherwise empty PR. + +## Risks + +PR 4a remains the largest necessary compatibility change: Core's service lifetime and configuration model cross repository package boundaries. The minimum slice is based on source inspection, not a completed migration experiment. Focused tests may identify additional inseparable changes. + +Intermediate Loader, Include and HMR copies intentionally combine an old baseline with documented upstream slices. Missing slice records would make a later sync unsafe. Deferred config, candidate validation, journal writes and asynchronous disposal must be tested in combination rather than inferred from independently passing unit cases. A new Include's type dependency on HMR must not force both concrete providers into the same incompatible `Context.hmr` declaration. From 4dee41f4bc71537d9e6c73c9f98cb4fa10253338 Mon Sep 17 00:00:00 2001 From: 07akioni <07akioni2@gmail.com> Date: Wed, 14 Oct 2026 15:23:58 +0800 Subject: [PATCH 024/162] feat(telemetry): attach cached desktop environment to product events --- docs/config-catalog.md | 6 +- docs/config-catalog.zh.md | 6 +- docs/subsystems/product-telemetry.md | 4 +- docs/subsystems/product-telemetry.zh.md | 4 +- .../web-app/tests/product-analytics.spec.ts | 6 +- .../client/product-analytics/README.i18n.yaml | 8 +- packages/client/product-analytics/README.md | 24 ++- .../client/product-analytics/README.zh.md | 24 ++- .../client/product-analytics/package.json | 4 +- .../src/environment-query.ts | 42 ++++ .../product-analytics/src/environment.ts | 172 +++++++++++++++++ .../client/product-analytics/src/index.ts | 35 +++- .../tests/collection.host.spec.ts | 74 +++++++- .../tests/environment-composition.host.e2e.ts | 43 +++++ .../tests/environment-query.host.spec.ts | 42 ++++ .../tests/environment.host.spec.ts | 179 ++++++++++++++++++ .../tests/fixtures/environment.host.ts | 66 +++++++ .../tests/fixtures/environment.patch.yml | 19 ++ .../product-analytics/tsconfig.host.json | 4 +- .../extensions/tool-cordis/src/api-catalog.ts | 2 +- pnpm-lock.yaml | 6 + tsconfig.client.json | 4 +- 22 files changed, 743 insertions(+), 31 deletions(-) create mode 100644 packages/client/product-analytics/src/environment-query.ts create mode 100644 packages/client/product-analytics/src/environment.ts create mode 100644 packages/client/product-analytics/tests/environment-composition.host.e2e.ts create mode 100644 packages/client/product-analytics/tests/environment-query.host.spec.ts create mode 100644 packages/client/product-analytics/tests/environment.host.spec.ts create mode 100644 packages/client/product-analytics/tests/fixtures/environment.host.ts create mode 100644 packages/client/product-analytics/tests/fixtures/environment.patch.yml diff --git a/docs/config-catalog.md b/docs/config-catalog.md index 8bfae862ef..c1d5e8e0f5 100644 --- a/docs/config-catalog.md +++ b/docs/config-catalog.md @@ -498,7 +498,7 @@ export interface Config { - `inject`: `deepseekAccount` · `productTelemetry` - `refs`: `Volatile` (`@deepseek-ai/cordis`) -- `source`: [`packages/client/product-analytics/src/index.ts:14`](../packages/client/product-analytics/src/index.ts) +- `source`: [`packages/client/product-analytics/src/index.ts:15`](../packages/client/product-analytics/src/index.ts) ```ts config-catalog /** Application-owned collection policy; no user settings surface. */ @@ -507,6 +507,10 @@ export interface Config { enabled: Volatile /** Running Desktop release, absent when unavailable. */ appVersion?: string + /** Maximum execution time per static environment query, in milliseconds. */ + environmentQueryTimeoutMillis: number + /** Maximum captured bytes per environment query output stream. */ + environmentQueryMaxOutputBytes: number } ``` diff --git a/docs/config-catalog.zh.md b/docs/config-catalog.zh.md index 68dfbf1852..068fed6c49 100644 --- a/docs/config-catalog.zh.md +++ b/docs/config-catalog.zh.md @@ -500,7 +500,7 @@ export interface Config { - `inject`: `deepseekAccount` · `productTelemetry` - `refs`: `Volatile` (`@deepseek-ai/cordis`) -- `source`: [`packages/client/product-analytics/src/index.ts:14`](../packages/client/product-analytics/src/index.ts) +- `source`: [`packages/client/product-analytics/src/index.ts:15`](../packages/client/product-analytics/src/index.ts) ```ts config-catalog /** Application-owned collection policy; no user settings surface. */ @@ -509,6 +509,10 @@ export interface Config { enabled: Volatile /** Running Desktop release, absent when unavailable. */ appVersion?: string + /** Maximum execution time per static environment query, in milliseconds. */ + environmentQueryTimeoutMillis: number + /** Maximum captured bytes per environment query output stream. */ + environmentQueryMaxOutputBytes: number } ``` diff --git a/docs/subsystems/product-telemetry.md b/docs/subsystems/product-telemetry.md index 00c70a45c8..9d656631f6 100644 --- a/docs/subsystems/product-telemetry.md +++ b/docs/subsystems/product-telemetry.md @@ -51,7 +51,9 @@ Authenticated event intake; disabled instances do not inspect identity or accept @Remote({ mode: 'stream' }) async *watchPolicy(signal: AbortSignal): AsyncIterable /** - * Submit selected Desktop fields; missing identity is omitted and never generated. + * Submit selected Desktop fields with cached static environment and fresh identity. + * The first intake waits for bounded environment queries; concurrent intakes share them. + * Missing identity is omitted; failed environment fields carry fixed reasons. * @param event - typed product event without message contents or credentials. * @returns after local submission; no delivery or warehouse acknowledgement. */ diff --git a/docs/subsystems/product-telemetry.zh.md b/docs/subsystems/product-telemetry.zh.md index 5b04bf6fdd..e8072807ec 100644 --- a/docs/subsystems/product-telemetry.zh.md +++ b/docs/subsystems/product-telemetry.zh.md @@ -51,7 +51,9 @@ Authenticated event intake; disabled instances do not inspect identity or accept @Remote({ mode: 'stream' }) async *watchPolicy(signal: AbortSignal): AsyncIterable /** - * Submit selected Desktop fields; missing identity is omitted and never generated. + * Submit selected Desktop fields with cached static environment and fresh identity. + * The first intake waits for bounded environment queries; concurrent intakes share them. + * Missing identity is omitted; failed environment fields carry fixed reasons. * @param event - typed product event without message contents or credentials. * @returns after local submission; no delivery or warehouse acknowledgement. */ diff --git a/packages/bundle/web-app/tests/product-analytics.spec.ts b/packages/bundle/web-app/tests/product-analytics.spec.ts index f90f7fade4..d9bf9f31dd 100644 --- a/packages/bundle/web-app/tests/product-analytics.spec.ts +++ b/packages/bundle/web-app/tests/product-analytics.spec.ts @@ -72,7 +72,11 @@ it.each(['desktop', 'web'])('limits collection and its shutdown to the Desktop l } const emit = vi.spyOn(ctx.productTelemetry, 'emit') await analytics.report({ eventName: 'desktop_app_launch', timestamp: 1, attributes: {} }) - expect(emit).toHaveBeenCalledWith(expect.objectContaining({ attributes: { app_version: 'test-version' } })) + expect(emit).toHaveBeenCalledOnce() + const attributes = emit.mock.calls[0]?.[0].attributes + expect(attributes?.app_version).toBe('test-version') + expect(typeof attributes?.os_name).toBe('string') + expect(typeof attributes?.memory_total_bytes).toBe('number') lifetime.abort() await policy.return?.() ctx.productTelemetry.emit({ eventName: 'desktop_upgrade_install_restart_click', body: 'upgrade', timestamp: Date.now() }) diff --git a/packages/client/product-analytics/README.i18n.yaml b/packages/client/product-analytics/README.i18n.yaml index 9b47a9f66e..989c03c5b2 100644 --- a/packages/client/product-analytics/README.i18n.yaml +++ b/packages/client/product-analytics/README.i18n.yaml @@ -15,11 +15,11 @@ en: 2fee9973f19f9401 zh: 8ab4b55413cb8408 /deepseek-ai-dsh-client-product-analytics/use-this-package: - en: 3932aceab765ab96 - zh: 8b5174fff1208053 + en: 062e1847cfcc60ee + zh: ea1b6cc2d2413226 /deepseek-ai-dsh-client-product-analytics/understand-the-implementation: - en: aa37aaf70a94e620 - zh: 23b9a01ed88cb772 + en: 96b1169d78f4c761 + zh: 475df4ea6ed1cb06 /deepseek-ai-dsh-client-product-analytics/model-experience: en: 291751b16b8a37e7 zh: 3b38008f4364ec41 diff --git a/packages/client/product-analytics/README.md b/packages/client/product-analytics/README.md index 941025b2de..1317bcd5fe 100644 --- a/packages/client/product-analytics/README.md +++ b/packages/client/product-analytics/README.md @@ -21,11 +21,27 @@ Desktop reports selected interactions through the existing OTel product exporter ## Use this package -Desktop mounts Analytics and its required Telemetry exporter. The `product-analytics` settings namespace owns the live `enabled` field, which defaults to `true`; configure it through the existing Cordis Config / settings mechanism. There is no user-facing control. Ordinary Web mounts neither service. Disabled collection reads no analytics identity and accepts no new events; the exporter remains mounted and may finish exporting records already queued. Session-feedback telemetry has its own policy. +Desktop mounts Analytics and its required Telemetry exporter. The `product-analytics` settings namespace owns the live `enabled` field, which defaults to `true`; configure it through the existing Cordis Config / settings mechanism. There is no user-facing control. Ordinary Web mounts neither service. Disabled collection starts no environment queries, reads no analytics identity and accepts no new events; the exporter remains mounted and may finish exporting records already queued. Session-feedback telemetry has its own policy. -The renderer and Electron subscribe to the Host policy through the existing authenticated stream, including reconnection. Electron also reads the initial policy before native launch reporting. Welcome obtains the current policy over IPC. The Host checks its current volatile configuration again at event intake and after identity lookup. `DSH_PRODUCT_ANALYTICS_OTLP_URL` overrides the export destination for isolated collectors. The [exporter](../../host/product-telemetry-otel/README.md) owns batching, retry, and shutdown delivery. +The renderer and Electron subscribe to the Host policy through the existing authenticated stream, including reconnection. Electron also reads the initial policy before native launch reporting. Welcome obtains the current policy over IPC. The Host checks its current volatile configuration again at event intake and after environment and identity lookup. `DSH_PRODUCT_ANALYTICS_OTLP_URL` overrides the export destination for isolated collectors. The [exporter](../../host/product-telemetry-otel/README.md) owns batching, retry, and shutdown delivery. -Common fields are `device_id`, `user_id`, `os_version`, and `app_version`. Device identity reuses the existing login record without generating one. The Host reads credential-free device, account, and OS fields through `deepseekAccount.getDeviceIdentity()`. Missing values are omitted; API keys, account tokens, prompts, and responses are never event fields. +Every product event carries available `device_id`, `user_id`, and `app_version` plus the static fields below. Identity reuses the existing login record without creating an identifier and is read for each event through `deepseekAccount.getDeviceIdentity()`. Static environment collection is independent of identity lookup. No separate environment event is emitted; API keys, account tokens, prompts, responses, hostnames, and machine UUIDs are excluded. + +| Field | Meaning and source | +|---|---| +| `os_name` | `windows` or `macos`; Node platform selects the mapping | +| `os_version` | Legacy ` ` from `node:os.release()` | +| `os_version_full` | Windows registry major/minor/build plus `UBR` revision; macOS `sw_vers` product version including a security-response suffix when present | +| `os_build` | Windows registry `CurrentBuildNumber`; macOS `sw_vers` build | +| `os_arch` | Native architecture mapped to `ia32`, `x64`, `arm`, or `arm64`; Windows processor CIM architecture or macOS `hw.optional.arm64`, independent of application emulation | +| `app_arch` | Running binary architecture from `process.arch` | +| `cpu_model` | First logical processor's model from `node:os.cpus()` | +| `cpu_logical_count` | Number of system-reported logical processors from `node:os.cpus()`, not application parallelism | +| `cpu_physical_core_count` | Sum of Windows processor CIM `NumberOfCores` across sockets; macOS `hw.physicalcpu` | +| `memory_total_bytes` | Total physical memory from `node:os.totalmem()`, in bytes without rounding | +| `missing_fields` | Object mapping omitted environment field names to `unavailable`, `invalid-value`, `query-failed`, `timeout`, `output-limit`, or `unsupported`; empty when all fields are available | + +The first enabled intake collects one environment snapshot; concurrent and later intakes reuse its successful fields and failures until the plugin remounts or the Host restarts. Intake waits for sampling while callers submit asynchronously; startup does not await collection. Identity changes remain visible on subsequent events. Each metadata command has a configurable `environmentQueryTimeoutMillis` deadline (5,000 ms) and `environmentQueryMaxOutputBytes` bound per output stream (16,384 bytes). Disposal cancels commands and waits for child closure; disabled or disposed intake never emits late results. Missing revision never becomes `.0`, and missing numeric fields never become zero. Only Windows and macOS have full-version and physical-core queries; other platforms mark those fields `unsupported`. Electron forwards its build-inlined `DSH_CLIENT_VERSION` to the Host; analytics and the exporter reuse this client version. The exporter requires it. The [Desktop composition](../../bundle/web-app/README.md) owns batching and timeout settings, including cancellation at the shutdown deadline. @@ -39,7 +55,7 @@ Only `plugin_toggle` carries `plugin_type`: plugin rows use `plugin`, and packag
Implementation internals — click to expand -Presentational components receive callbacks. Shared Client services resolve the optional Desktop analytics sender, which requires both the native preload marker and the synchronized Host policy. The sender retains its own RPC context when called by other plugins. Native actions use the authenticated Host API; its generated Typert validator accepts only the typed event fields. The Host enriches identity immediately before queueing. It observes live compaction events without replaying Session history. +Presentational components receive callbacks. Shared Client services resolve the optional Desktop analytics sender, which requires both the native preload marker and the synchronized Host policy. The sender retains its own RPC context when called by other plugins. Native actions use the authenticated Host API; its generated Typert validator accepts only the typed event fields. The Host attaches its cached static environment and current identity before queueing. It observes live compaction events without replaying Session history.
diff --git a/packages/client/product-analytics/README.zh.md b/packages/client/product-analytics/README.zh.md index be44c8e742..c5d588b851 100644 --- a/packages/client/product-analytics/README.zh.md +++ b/packages/client/product-analytics/README.zh.md @@ -21,11 +21,27 @@ kind: "package-reference" ## 使用本包 -桌面端同时装配 Analytics 与必需的 Telemetry 导出器。`product-analytics` settings 命名空间持有动态 `enabled` 字段，默认 `true`，通过现有 Cordis Config / settings 机制配置，暂不提供用户操作入口。普通 Web 不装配这两个服务。关闭后不读取埋点身份、不接收新事件；导出器仍保持挂载，已入队事件可以继续导出。会话反馈遥测采用独立策略。 +桌面端同时装配 Analytics 与必需的 Telemetry 导出器。`product-analytics` settings 命名空间持有动态 `enabled` 字段，默认 `true`，通过现有 Cordis Config / settings 机制配置，暂不提供用户操作入口。普通 Web 不装配这两个服务。关闭后不启动环境查询、不读取埋点身份、不接收新事件；导出器仍保持挂载，已入队事件可以继续导出。会话反馈遥测采用独立策略。 -渲染端和 Electron 通过现有认证流订阅 Host 策略变化及重连后的值；Electron 在原生启动上报前还会读取初始策略。欢迎窗口通过 IPC 获取当前策略。Host 在接收事件及读取身份后都检查当前 volatile 配置。`DSH_PRODUCT_ANALYTICS_OTLP_URL` 可覆盖导出目的地，用于隔离的接收端。[导出器](../../host/product-telemetry-otel/README.zh.md)负责批量发送、重试和退出时的交付。 +渲染端和 Electron 通过现有认证流订阅 Host 策略变化及重连后的值；Electron 在原生启动上报前还会读取初始策略。欢迎窗口通过 IPC 获取当前策略。Host 在接收事件及读取环境和身份后都检查当前 volatile 配置。`DSH_PRODUCT_ANALYTICS_OTLP_URL` 可覆盖导出目的地，用于隔离的接收端。[导出器](../../host/product-telemetry-otel/README.zh.md)负责批量发送、重试和退出时的交付。 -公共字段为 `device_id`、`user_id`、`os_version` 和 `app_version`。设备身份复用现有登录记录，不会生成新标识。Host 通过 `deepseekAccount.getDeviceIdentity()` 读取不含凭据的设备、账户和操作系统字段。缺失值会省略；API key、账户令牌、提示词和模型回复都不是事件字段。 +每条产品事件携带可用的 `device_id`、`user_id` 和 `app_version`，以及下表中的静态字段。身份复用现有登录记录，不创建设备标识；每条事件通过 `deepseekAccount.getDeviceIdentity()` 读取身份。静态环境采集独立于身份读取。不发送单独的环境事件；排除 API key、账户令牌、提示词、模型回复、主机名和机器 UUID。 + +| 字段 | 含义与来源 | +|---|---| +| `os_name` | `windows` 或 `macos`；由 Node 平台选择映射 | +| `os_version` | 原有的 ` <内核版本>`，来自 `node:os.release()` | +| `os_version_full` | Windows 注册表 major/minor/build 加上 `UBR` 修订号；macOS `sw_vers` 产品版本，包括存在时的安全响应后缀 | +| `os_build` | Windows 注册表 `CurrentBuildNumber`；macOS `sw_vers` build | +| `os_arch` | 原生架构映射为 `ia32`、`x64`、`arm` 或 `arm64`；来自 Windows 处理器 CIM 架构或 macOS `hw.optional.arm64`，独立于应用仿真运行 | +| `app_arch` | 来自 `process.arch` 的运行二进制架构 | +| `cpu_model` | 来自 `node:os.cpus()` 的首个逻辑处理器型号 | +| `cpu_logical_count` | 来自 `node:os.cpus()` 的系统报告逻辑处理器数量，不表示应用可用并行度 | +| `cpu_physical_core_count` | Windows 各 socket 处理器 CIM `NumberOfCores` 之和；macOS `hw.physicalcpu` | +| `memory_total_bytes` | 来自 `node:os.totalmem()` 的总物理内存，单位 bytes，不取整 | +| `missing_fields` | 对象，将省略的环境字段名映射到 `unavailable`、`invalid-value`、`query-failed`、`timeout`、`output-limit` 或 `unsupported`；全部字段可用时为空 | + +首次启用状态下的事件接收采集一份环境快照；并发和后续事件复用成功字段及失败状态，直到插件重新挂载或 Host 重启。事件接收等待采样，调用方异步提交；启动不等待采集。后续事件仍能反映身份变化。每条元数据命令具有可配置的 `environmentQueryTimeoutMillis` 期限（5,000 毫秒）和每路输出的 `environmentQueryMaxOutputBytes` 上限（16,384 bytes）。卸载取消命令并等待子进程关闭；禁用或卸载后的事件接收不会发送迟到结果。缺失修订号不会变为 `.0`，缺失数字字段不会变为零。只有 Windows 和 macOS 查询完整版本与物理核心数；其他平台将这些字段标记为 `unsupported`。 Electron 将构建内联的 `DSH_CLIENT_VERSION` 传给 Host；埋点与导出器复用这一客户端版本，导出器要求该值存在。[桌面组合](../../bundle/web-app/README.zh.md)负责批量发送与超时配置，包括达到关闭期限时的取消行为。 @@ -39,7 +55,7 @@ Electron 将构建内联的 `DSH_CLIENT_VERSION` 传给 Host；埋点与导出
实现细节——点击展开 -展示组件接收回调。共享 Client 服务读取可选的桌面埋点发送器，后者要求同时存在原生 preload 标识和 同步的 Host 采集策略。发送器在被其他插件调用时保留自身的 RPC 上下文。原生操作使用已认证的 Host API，其生成的 Typert 校验器只接受类型中声明的事件字段。Host 在入队前补充身份信息，并监听实时压缩事件，不重放会话历史。 +展示组件接收回调。共享 Client 服务读取可选的桌面埋点发送器，后者要求同时存在原生 preload 标识和 同步的 Host 采集策略。发送器在被其他插件调用时保留自身的 RPC 上下文。原生操作使用已认证的 Host API，其生成的 Typert 校验器只接受类型中声明的事件字段。Host 在入队前附加缓存的静态环境和当前身份信息，并监听实时压缩事件，不重放会话历史。
diff --git a/packages/client/product-analytics/package.json b/packages/client/product-analytics/package.json index a8240ccc6b..02c111a9b7 100644 --- a/packages/client/product-analytics/package.json +++ b/packages/client/product-analytics/package.json @@ -64,7 +64,9 @@ "@deepseek-ai/dsh-host-webserver": "workspace:*", "@deepseek-ai/dsh-settings": "workspace:*", "@deepseek-ai/cordis-plugin-loader": "workspace:~", - "@deepseek-ai/dsh-otel": "workspace:*" + "@deepseek-ai/dsh-otel": "workspace:*", + "@deepseek-ai/dsh-app-boot": "workspace:*", + "@deepseek-ai/dsh-loader-smoke": "workspace:*" }, "files": [ "locale/*.json", diff --git a/packages/client/product-analytics/src/environment-query.ts b/packages/client/product-analytics/src/environment-query.ts new file mode 100644 index 0000000000..8086fff82f --- /dev/null +++ b/packages/client/product-analytics/src/environment-query.ts @@ -0,0 +1,42 @@ +/** Bounded, cancellable queries of fixed operating-system metadata commands. */ +import { execFile } from 'node:child_process' + +/** Fixed reasons exposed instead of command output or raw errors. */ +export type MissingReason = 'unavailable' | 'invalid-value' | 'query-failed' | 'timeout' | 'output-limit' | 'unsupported' + +/** Successful stdout or an isolated query failure. */ +export type QueryResult = { kind: 'value'; stdout: string } | { kind: 'missing'; reason: MissingReason } + +/** Deployment limits and the owning analytics lifetime. */ +export interface QueryOptions { + /** Maximum execution time of each query, in milliseconds. */ + timeoutMillis: number + /** Maximum bytes captured from either output stream. */ + maxOutputBytes: number + /** Disposal aborts the command and waits for its close event. */ + signal: AbortSignal +} + +/** + * Execute fixed argv without a shell; failures omit only dependent metadata. + * @param command - operating-system executable, never event-supplied text. + * @param args - fixed metadata query arguments. + * @param options - output/time limits and owner lifetime. + * @returns stdout or a fixed failure reason, after the child closes. + */ +export async function queryEnvironment(command: string, args: readonly string[], options: QueryOptions): Promise { + const result = Promise.withResolvers() + const child = execFile(command, [...args], { + encoding: 'utf8', windowsHide: true, signal: options.signal, killSignal: 'SIGKILL', + timeout: options.timeoutMillis, maxBuffer: options.maxOutputBytes, + }, (error, stdout) => { + if (error === null) { result.resolve({ kind: 'value', stdout }); return } + const reason = error.code === 'ERR_CHILD_PROCESS_STDIO_MAXBUFFER' ? 'output-limit' + : error.killed && !options.signal.aborted ? 'timeout' : 'query-failed' + result.resolve({ kind: 'missing', reason }) + }) + // execFile can report abort before stdio closes; disposal owns both settlements. + const closed = new Promise((resolve) => { child.once('close', () => { resolve() }) }) + const [output] = await Promise.all([result.promise, closed]) + return output +} diff --git a/packages/client/product-analytics/src/environment.ts b/packages/client/product-analytics/src/environment.ts new file mode 100644 index 0000000000..7f067cc7b8 --- /dev/null +++ b/packages/client/product-analytics/src/environment.ts @@ -0,0 +1,172 @@ +/** Static Desktop environment fields, independent of account and credential availability. */ +import { cpus, machine, platform, release, totalmem } from 'node:os' +import { join } from 'node:path' +import { z } from 'zod' +import { queryEnvironment, type MissingReason, type QueryOptions, type QueryResult } from './environment-query.ts' + +/** Approved scalar attributes; missing fields carry fixed reasons. */ +export interface EnvironmentAttributes { + /** Platform name: windows, macos, or the unsupported Node platform. */ + os_name: string + /** Legacy platform and kernel release string. */ + os_version?: string + /** Windows major.minor.build.revision or macOS product version. */ + os_version_full?: string + /** Windows build or macOS product build. */ + os_build?: string + /** Native system architecture, independent of process emulation. */ + os_arch?: string + /** Architecture of the running application binary. */ + app_arch: string + /** Model reported by the first logical processor. */ + cpu_model?: string + /** System-reported logical processor count, not available parallelism. */ + cpu_logical_count?: number + /** Total physical cores across CPU sockets; never inferred from logical processors. */ + cpu_physical_core_count?: number + /** Total physical memory in bytes. */ + memory_total_bytes?: number + /** Only failed fields, with no raw commands, output, or errors. */ + missing_fields: Record +} + +type EnvironmentField = Exclude + +const windowsVersion = z.object({ + major: z.number().int().nonnegative().nullable(), minor: z.number().int().nonnegative().nullable(), + build: z.string().regex(/^\d+$/).nullable(), revision: z.number().int().nonnegative().nullable(), +}) +const windowsCpu = z.array(z.object({ cores: z.number().int().positive().nullable(), architecture: z.number().int().nullable() })).min(1) + +const VERSION_QUERY = ` +$ErrorActionPreference = 'Stop' +$r = Get-Item -LiteralPath 'HKLM:\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion' +@{ major = $r.GetValue('CurrentMajorVersionNumber'); minor = $r.GetValue('CurrentMinorVersionNumber'); build = $r.GetValue('CurrentBuildNumber'); revision = $r.GetValue('UBR') } | ConvertTo-Json -Compress +` +const CPU_QUERY = ` +$ErrorActionPreference = 'Stop' +ConvertTo-Json -Compress -InputObject @(Get-CimInstance -ClassName Win32_Processor -Property NumberOfCores,Architecture | ForEach-Object { @{ cores = $_.NumberOfCores; architecture = $_.Architecture } }) +` + +/** + * Collect only approved metadata; a failed source leaves its dependent fields missing. + * @param options - per-command bounds and analytics disposal signal. + * @returns a snapshot for reuse throughout this analytics instance, including failed fields. + */ +export async function collectEnvironment(options: QueryOptions): Promise { + const hostPlatform = platform() + const fields: EnvironmentAttributes = { + os_name: hostPlatform === 'darwin' ? 'macos' : hostPlatform === 'win32' ? 'windows' : hostPlatform, + app_arch: process.arch, missing_fields: {}, + } + const missing = (keys: readonly EnvironmentField[], reason: MissingReason): void => { + for (const key of keys) fields.missing_fields[key] = reason + } + const read = (key: K, source: () => EnvironmentAttributes[K]): void => { + let value: EnvironmentAttributes[K] + try { value = source() } catch (_error) { + // The failed OS source cannot invalidate independently collected fields. + missing([key], 'unavailable'); return + } + if (value === undefined || value === '' || (typeof value === 'number' && (!Number.isSafeInteger(value) || value <= 0))) { + missing([key], 'unavailable'); return + } + Object.assign(fields, { [key]: value }) + } + read('os_version', () => `${hostPlatform} ${release()}`) + read('memory_total_bytes', totalmem) + let processors: ReturnType = [] + try { processors = cpus() } catch (_error) { + // CPU metadata is optional; memory and OS collection remain usable. + } + read('cpu_model', () => processors[0]?.model.trim()) + read('cpu_logical_count', () => processors.length) + + if (hostPlatform === 'win32') { + const root = process.env.SystemRoot + if (root === undefined || root === '') { + missing(['os_version_full', 'os_build', 'os_arch', 'cpu_physical_core_count'], 'unavailable') + return fields + } + const command = join(root, 'System32', 'WindowsPowerShell', 'v1.0', 'powershell.exe') + const query = (script: string): Promise => queryEnvironment(command, + ['-NoProfile', '-NonInteractive', '-EncodedCommand', Buffer.from(script, 'utf16le').toString('base64')], options) + const [version, cpu] = await Promise.all([query(VERSION_QUERY), query(CPU_QUERY)]) + if (version.kind === 'missing') missing(['os_version_full', 'os_build'], version.reason) + else { + const parsed = decode(version.stdout, windowsVersion) + if (parsed === undefined) missing(['os_version_full', 'os_build'], 'invalid-value') + else { + if (parsed.build === null) missing(['os_build'], 'unavailable') + else fields.os_build = parsed.build + if (parsed.major === null || parsed.minor === null || parsed.build === null || parsed.revision === null) missing(['os_version_full'], 'unavailable') + else fields.os_version_full = `${parsed.major}.${parsed.minor}.${parsed.build}.${parsed.revision}` + } + } + if (cpu.kind === 'missing') missing(['os_arch', 'cpu_physical_core_count'], cpu.reason) + else { + const parsed = decode(cpu.stdout, windowsCpu) + if (parsed === undefined) missing(['os_arch', 'cpu_physical_core_count'], 'invalid-value') + else { + const counts = parsed.map(value => value.cores) + if (counts.every((count): count is number => count !== null)) read('cpu_physical_core_count', () => counts.reduce((sum, count) => sum + count, 0)) + else missing(['cpu_physical_core_count'], 'unavailable') + const architectures: Record = { 0: 'ia32', 5: 'arm', 9: 'x64', 12: 'arm64' } + const arch = parsed[0]?.architecture + if (arch == null || !parsed.every(value => value.architecture === arch) || architectures[arch] === undefined) missing(['os_arch'], 'unavailable') + else fields.os_arch = architectures[arch] + } + } + } else if (hostPlatform === 'darwin') { + const [version, cpu] = await Promise.all([ + queryEnvironment('/usr/bin/sw_vers', [], options), + queryEnvironment('/usr/sbin/sysctl', ['-i', 'hw.physicalcpu', 'hw.optional.arm64'], options), + ]) + if (version.kind === 'missing') missing(['os_version_full', 'os_build'], version.reason) + else { + const values = lines(version.stdout) + const product = values.get('ProductVersion') + const extra = values.get('ProductVersionExtra') + if (product !== undefined && /^\d+(?:\.\d+){1,2}$/.test(product) && (extra === undefined || /^\([a-z]\)$/.test(extra))) { + fields.os_version_full = extra === undefined ? product : `${product} ${extra}` + } else missing(['os_version_full'], 'invalid-value') + const build = values.get('BuildVersion') + if (build !== undefined && /^\d+[A-Za-z]\d+[a-z]?$/.test(build)) fields.os_build = build + else missing(['os_build'], 'invalid-value') + } + if (cpu.kind === 'missing') missing(['os_arch', 'cpu_physical_core_count'], cpu.reason) + else { + const values = lines(cpu.stdout) + const cores = values.get('hw.physicalcpu') + if (cores !== undefined && /^\d+$/.test(cores) && Number.isSafeInteger(Number(cores)) && Number(cores) > 0) { + fields.cpu_physical_core_count = Number(cores) + } + else missing(['cpu_physical_core_count'], 'invalid-value') + const arm = values.get('hw.optional.arm64') + if (arm === '1') fields.os_arch = 'arm64' + else if (arm === '0') fields.os_arch = 'x64' + else missing(['os_arch'], 'unavailable') + } + } else { + read('os_arch', () => machine()) + missing(['os_version_full', 'os_build', 'cpu_physical_core_count'], 'unsupported') + } + return fields +} + +function lines(stdout: string): Map { + return new Map(stdout.trim().split(/\r?\n/).map((line) => { + const colon = line.indexOf(':') + return [line.slice(0, colon).trim(), line.slice(colon + 1).trim()] + })) +} + +function decode(stdout: string, schema: z.ZodType): T | undefined { + let value: unknown + try { value = JSON.parse(stdout) } catch (_error) { + // Malformed command JSON supplies no validated metadata. + return undefined + } + const result = schema.safeParse(value) + return result.success ? result.data : undefined +} diff --git a/packages/client/product-analytics/src/index.ts b/packages/client/product-analytics/src/index.ts index 3eae234d57..ce497f99df 100644 --- a/packages/client/product-analytics/src/index.ts +++ b/packages/client/product-analytics/src/index.ts @@ -9,6 +9,7 @@ import type {} from '@deepseek-ai/cordis-plugin-loader' import type {} from '@deepseek-ai/dsh-session' import type {} from '@deepseek-ai/dsh-compaction/types' import type { ProductEvent } from './events.ts' +import { collectEnvironment, type EnvironmentAttributes } from './environment.ts' /** Application-owned collection policy; no user settings surface. */ export interface Config { @@ -16,6 +17,10 @@ export interface Config { enabled: Volatile /** Running Desktop release, absent when unavailable. */ appVersion?: string + /** Maximum execution time per static environment query, in milliseconds. */ + environmentQueryTimeoutMillis: number + /** Maximum captured bytes per environment query output stream. */ + environmentQueryMaxOutputBytes: number } declare module '@deepseek-ai/cordis' { @@ -27,13 +32,24 @@ declare module '@deepseek-ai/cordis' { /** Authenticated event intake; disabled instances do not inspect identity or accept new events. */ export default class ProductAnalytics extends TypertRemoteService { static inject = ['deepseekAccount', 'productTelemetry'] - static Config = z.object({ enabled: z.boolean().default(true).volatile(), appVersion: z.string() }) + static Config = z.object({ + enabled: z.boolean().default(true).volatile(), appVersion: z.string(), + environmentQueryTimeoutMillis: z.number().step(1).min(1).max(2_147_483_647).default(5000), + environmentQueryMaxOutputBytes: z.number().step(1).min(1).max(2_147_483_647).default(16384), + }) private active = true + private readonly environmentLifetime = new AbortController() + private environment: Promise | undefined private readonly listeners = new Set<() => void>() constructor(ctx: Context, private readonly config: Config) { super(ctx, 'productAnalytics') - ctx.effect(() => () => { this.active = false; for (const listener of this.listeners) listener() }) + ctx.effect(() => async () => { + this.active = false + for (const listener of this.listeners) listener() + this.environmentLifetime.abort() + await this.environment + }) ctx.on('loader/volatile-update', () => { for (const listener of this.listeners) listener() }) ctx.inject(['settings'], (child) => { child.effect(() => child.settings.configure({ auto: false }, ctx.fiber)) }) ctx.on('session/event', (session, event) => { @@ -74,7 +90,9 @@ export default class ProductAnalytics extends TypertRemoteService { } /** - * Submit selected Desktop fields; missing identity is omitted and never generated. + * Submit selected Desktop fields with cached static environment and fresh identity. + * The first intake waits for bounded environment queries; concurrent intakes share them. + * Missing identity is omitted; failed environment fields carry fixed reasons. * @param event - typed product event without message contents or credentials. * @returns after local submission; no delivery or warehouse acknowledgement. */ @@ -82,16 +100,23 @@ export default class ProductAnalytics extends TypertRemoteService { async report(event: ProductEvent): Promise { if (!this.enabled()) return try { - const identity = await this.ctx.deepseekAccount.getDeviceIdentity().catch(() => undefined) + this.environment ??= collectEnvironment({ + timeoutMillis: this.config.environmentQueryTimeoutMillis, + maxOutputBytes: this.config.environmentQueryMaxOutputBytes, + signal: this.environmentLifetime.signal, + }) + const [environment, identity] = await Promise.all([ + this.environment, this.ctx.deepseekAccount.getDeviceIdentity().catch(() => undefined), + ]) if (!this.enabled()) return this.ctx.productTelemetry.emit({ ...event, body: event.eventName, attributes: { ...event.attributes, + ...environment, ...identity?.deviceId === undefined ? {} : { device_id: identity.deviceId }, ...identity?.userId === undefined ? {} : { user_id: identity.userId }, ...this.config.appVersion === undefined ? {} : { app_version: this.config.appVersion }, - ...identity === undefined ? {} : { os_version: identity.osVersion }, }, }) } catch (error) { diff --git a/packages/client/product-analytics/tests/collection.host.spec.ts b/packages/client/product-analytics/tests/collection.host.spec.ts index c7878a8940..5538f4c513 100644 --- a/packages/client/product-analytics/tests/collection.host.spec.ts +++ b/packages/client/product-analytics/tests/collection.host.spec.ts @@ -2,11 +2,20 @@ import OTel from '@deepseek-ai/dsh-otel' import { Context } from '@deepseek-ai/cordis' import { createServer } from 'node:http' import { once } from 'node:events' -import { afterEach, expect, it, vi } from 'vitest' +import { afterEach, beforeEach, expect, it, vi } from 'vitest' import ProductTelemetry, { Config as TelemetryConfig, type ProductTelemetryRecord } from '@deepseek-ai/dsh-host-product-telemetry-otel' import { Session, SessionId, SessionSeq } from '@deepseek-ai/dsh-session' import { CompactionId } from '@deepseek-ai/dsh-compaction' import Analytics from '../src/index.ts' +import { collectEnvironment, type EnvironmentAttributes } from '../src/environment.ts' + +vi.mock('../src/environment.ts', () => ({ collectEnvironment: vi.fn() })) +const environment: EnvironmentAttributes = { + os_name: 'windows', os_version: 'win32 10.0.26300', os_version_full: '10.0.26300.9457', os_build: '26300', + os_arch: 'arm64', app_arch: 'x64', cpu_model: 'Fixture CPU', cpu_logical_count: 16, cpu_physical_core_count: 8, + memory_total_bytes: 34359738368, missing_fields: {}, +} +beforeEach(() => { vi.resetAllMocks(); vi.mocked(collectEnvironment).mockResolvedValue(environment) }) const cleanup: (() => Promise)[] = [] afterEach(async () => { for (const dispose of cleanup.splice(0).reverse()) await dispose() }) @@ -27,6 +36,7 @@ it('disabled collection does not read identity or submit an event', async () => const b = await setup(false) expect(b.ctx.productAnalytics.enabled()).toBe(false) await b.ctx.productAnalytics.report({ eventName: 'desktop_app_launch', timestamp: 100, attributes: {} }) + expect(collectEnvironment).not.toHaveBeenCalled() expect(b.getDeviceIdentity).not.toHaveBeenCalled() expect(b.emit).not.toHaveBeenCalled() }) @@ -37,7 +47,7 @@ it('reuses login identity and copies only approved common fields', async () => { expect(b.getDeviceIdentity).toHaveBeenCalledWith() expect(b.emit).toHaveBeenCalledExactlyOnceWith({ eventName: 'auth_page_click', body: 'auth_page_click', timestamp: 100, - attributes: { button_name: 'sign_in', device_id: 'login-device', user_id: 'user-1', app_version: 'test-version', os_version: expect.any(String) as string }, + attributes: { ...environment, button_name: 'sign_in', device_id: 'login-device', user_id: 'user-1', app_version: 'test-version' }, }) expect(JSON.stringify(b.emit.mock.calls)).not.toContain('private-') }) @@ -46,7 +56,7 @@ it('missing identity does not discard an otherwise valid event', async () => { const b = await setup(true) b.getDeviceIdentity.mockRejectedValueOnce(new Error('credential store unavailable')) await b.ctx.productAnalytics.report({ eventName: 'plugin_add_button_click', timestamp: 100, attributes: {} }) - expect(b.emit.mock.calls[0]?.[0].attributes).toEqual({ app_version: 'test-version' }) + expect(b.emit.mock.calls[0]?.[0].attributes).toEqual({ ...environment, app_version: 'test-version' }) }) it('unload suppresses an identity lookup that settles after disposal', async () => { @@ -79,11 +89,21 @@ it('writes the selected event through the real exporter to an isolated collector ctx.provide('webServer', {} as never) await ctx.plugin(OTel) const exporter = await ctx.plugin(ProductTelemetry, TelemetryConfig({ endpoint: `http://127.0.0.1:${address.port}/v1/logs`, serviceName: 'test', serviceVersion: '1', compression: 'none' })) - await ctx.plugin(Analytics, { enabled: true }) + const analytics = await ctx.plugin(Analytics, { enabled: true }) await ctx.productAnalytics.report({ eventName: 'api_key_save_click', timestamp: 1_800_000_000_000, attributes: {} }) + await analytics.dispose() + vi.mocked(collectEnvironment).mockResolvedValueOnce({ ...environment, os_version_full: '10.0.26300.9458' }) + await ctx.plugin(Analytics, { enabled: true }) + await ctx.productAnalytics.report({ eventName: 'api_key_save_click', timestamp: 1_800_000_000_001, attributes: {} }) await exporter.dispose() expect(captures).toHaveLength(1) - expect(JSON.parse(captures[0]!)).toMatchObject({ resourceLogs: [{ scopeLogs: [{ logRecords: [{ eventName: 'api_key_save_click', timeUnixNano: '1800000000000000000' }] }] }] }) + expect(JSON.parse(captures[0]!)).toMatchObject({ resourceLogs: [{ scopeLogs: [{ logRecords: [ + { eventName: 'api_key_save_click', timeUnixNano: '1800000000000000000' }, + { eventName: 'api_key_save_click', timeUnixNano: '1800000000001000000' }, + ] }] }] }) + expect(captures[0]).toContain('10.0.26300.9457') + expect(captures[0]).toContain('10.0.26300.9458') + expect(collectEnvironment).toHaveBeenCalledTimes(2) expect(captures[0]).not.toContain('device_id') expect(captures[0]).not.toContain('app_version') }) @@ -159,3 +179,47 @@ it('wakes pending policy readers on disposal and registers hidden settings prese expect(await pending).toEqual({ done: true, value: undefined }) expect(dispose).toHaveBeenCalledOnce() }) + +it('shares one environment query across concurrent events and keeps identity fresh', async () => { + const b = await setup(true) + const pending = Promise.withResolvers() + vi.mocked(collectEnvironment).mockReturnValueOnce(pending.promise) + const first = b.ctx.productAnalytics.report({ eventName: 'auth_page_view', timestamp: 1, attributes: {} }) + b.getDeviceIdentity.mockResolvedValueOnce({ userId: 'user-2', osVersion: 'untrusted-os' }) + const second = b.ctx.productAnalytics.report({ eventName: 'desktop_app_launch', timestamp: 2, attributes: {} }) + expect(collectEnvironment).toHaveBeenCalledTimes(1) + expect(b.emit).not.toHaveBeenCalled() + pending.resolve(environment) + await Promise.all([first, second]) + expect(b.emit.mock.calls.map(([record]) => record.attributes?.user_id)).toEqual(['user-1', 'user-2']) + expect(b.emit.mock.calls.map(([record]) => record.attributes?.os_version_full)).toEqual(['10.0.26300.9457', '10.0.26300.9457']) + expect(b.emit.mock.calls).toHaveLength(2) + await b.ctx.productAnalytics.report({ eventName: 'auth_page_view', timestamp: 3, attributes: {} }) + expect(collectEnvironment).toHaveBeenCalledTimes(1) + expect(b.getDeviceIdentity).toHaveBeenCalledTimes(3) +}) + +it('caches failed environment fields without discarding the event or retrying each interaction', async () => { + const b = await setup(true) + vi.mocked(collectEnvironment).mockResolvedValue({ os_name: 'windows', app_arch: 'x64', missing_fields: { os_version_full: 'timeout' } }) + for (const timestamp of [1, 2]) await b.ctx.productAnalytics.report({ eventName: 'auth_page_view', timestamp, attributes: {} }) + expect(collectEnvironment).toHaveBeenCalledTimes(1) + expect(b.emit.mock.calls).toHaveLength(2) + expect(b.emit.mock.calls[0]![0].attributes?.missing_fields).toEqual({ os_version_full: 'timeout' }) +}) + +it('aborts and awaits environment sampling on disposal without emitting late results', async () => { + const b = await setup(true) + const pending = Promise.withResolvers() + vi.mocked(collectEnvironment).mockReturnValueOnce(pending.promise) + const report = b.ctx.productAnalytics.report({ eventName: 'desktop_app_launch', timestamp: 1, attributes: {} }) + const signal = vi.mocked(collectEnvironment).mock.calls[0]![0].signal + const aborted = new Promise((resolve) => { signal.addEventListener('abort', () => { resolve() }, { once: true }) }) + let disposed = false + const disposal = b.fiber.dispose().then(() => { disposed = true }) + await aborted + expect(disposed).toBe(false) + pending.resolve(environment) + await Promise.all([report, disposal]) + expect(b.emit).not.toHaveBeenCalled() +}) diff --git a/packages/client/product-analytics/tests/environment-composition.host.e2e.ts b/packages/client/product-analytics/tests/environment-composition.host.e2e.ts new file mode 100644 index 0000000000..6bb08a171a --- /dev/null +++ b/packages/client/product-analytics/tests/environment-composition.host.e2e.ts @@ -0,0 +1,43 @@ +import { readFile } from 'node:fs/promises' +import { resolve } from 'node:path' +import { platform } from 'node:os' +import { expect, it } from 'vitest' +import { z } from 'zod' +import { runLoaderSmoke, LOADER_SMOKE_TEST_TIMEOUT_MS } from '@deepseek-ai/dsh-loader-smoke' + +it('adds real system metadata to the existing event through Loader and OTLP', async () => { + const driver = resolve(import.meta.dirname, 'fixtures/environment.host.ts') + let captured: unknown + const attribute = z.object({ key: z.string(), value: z.object({ + stringValue: z.string().optional(), intValue: z.number().optional(), + kvlistValue: z.object({ values: z.array(z.unknown()) }).optional(), + }) }) + const record = z.object({ eventName: z.string(), attributes: z.array(attribute) }) + const batch = z.object({ resourceLogs: z.array(z.object({ scopeLogs: z.array(z.object({ logRecords: z.array(record) })) })) }) + const schema = z.object({ captures: z.array(batch), reference: z.record(z.string(), z.union([z.string(), z.number()])) }) + await runLoaderSmoke({ + label: 'Desktop environment attributes', tempDirPrefix: 'product-environment-', + binScript: driver, libBinScript: driver, + configPath: resolve(import.meta.dirname, 'fixtures/environment.patch.yml'), + tsconfigPath: resolve(import.meta.dirname, '../../../../tsconfig.base.json'), + inspect: async (cwd) => { captured = JSON.parse(await readFile(resolve(cwd, 'captures.json'), 'utf8')) }, + }) + const result = schema.parse(captured) + const records = result.captures.flatMap(batch => batch.resourceLogs + .flatMap(resource => resource.scopeLogs.flatMap(scope => scope.logRecords))) + expect(records).toHaveLength(1) + expect(records[0]!.eventName).toBe('desktop_app_launch') + const fields = Object.fromEntries(records[0]!.attributes.map(attribute => [attribute.key, attribute.value])) + for (const [key, value] of Object.entries(result.reference)) { + expect(fields[key]).toEqual(typeof value === 'number' ? { intValue: value } : { stringValue: value }) + } + if (platform() === 'win32' || platform() === 'darwin') { + expect(fields.missing_fields).toEqual({ kvlistValue: { values: [] } }) + expect(fields.cpu_model?.stringValue).toBeTruthy() + } + if (platform() === 'win32') expect(fields.os_version_full?.stringValue).toMatch(/^\d+\.\d+\.\d+\.\d+$/) + expect(fields.app_version).toEqual({ stringValue: 'synthetic-release' }) + expect(fields).not.toHaveProperty('device_id') + expect(fields).not.toHaveProperty('user_id') + expect(JSON.stringify(result.captures)).not.toMatch(/serial_number|hostname|device_uuid|environment_schema_version/) +}, LOADER_SMOKE_TEST_TIMEOUT_MS) diff --git a/packages/client/product-analytics/tests/environment-query.host.spec.ts b/packages/client/product-analytics/tests/environment-query.host.spec.ts new file mode 100644 index 0000000000..6b84b4ee76 --- /dev/null +++ b/packages/client/product-analytics/tests/environment-query.host.spec.ts @@ -0,0 +1,42 @@ +import { expect, it } from 'vitest' +import { queryEnvironment } from '../src/environment-query.ts' + +const options = () => ({ timeoutMillis: 5000, maxOutputBytes: 16384, signal: new AbortController().signal }) + +// These Node snippets are metadata-command stand-ins, with no Cordis or application boot. +it('captures UTF-8 output and closes the child before returning', async () => { + expect(await queryEnvironment(process.execPath, ['-e', "process.stdout.write('CPU 型号')"], options())) + .toEqual({ kind: 'value', stdout: 'CPU 型号' }) +}) + +it('omits nonzero command output and raw error text', async () => { + expect(await queryEnvironment(process.execPath, ['-e', "console.error('private error'); console.log('private output'); process.exitCode = 1"], options())) + .toEqual({ kind: 'missing', reason: 'query-failed' }) +}) + +it('settles a failed spawn without retaining an output buffer or child', async () => { + expect(await queryEnvironment(`${process.execPath}-nonexistent`, [], options())) + .toEqual({ kind: 'missing', reason: 'query-failed' }) +}) + +it('terminates a timed-out query and awaits close', async () => { + expect(await queryEnvironment(process.execPath, ['-e', 'setInterval(() => {}, 1000)'], { ...options(), timeoutMillis: 20 })) + .toEqual({ kind: 'missing', reason: 'timeout' }) +}) + +it('bounds multibyte stdout without treating truncated metadata as valid', async () => { + expect(await queryEnvironment(process.execPath, ['-e', "process.stdout.write('界'.repeat(10000))"], { ...options(), maxOutputBytes: 128 })) + .toEqual({ kind: 'missing', reason: 'output-limit' }) +}) + +it('bounds stderr even when stdout could describe valid metadata', async () => { + expect(await queryEnvironment(process.execPath, ['-e', "process.stderr.write('x'.repeat(10000))"], { ...options(), maxOutputBytes: 128 })) + .toEqual({ kind: 'missing', reason: 'output-limit' }) +}) + +it('aborts an in-flight query and waits for child closure', async () => { + const lifetime = new AbortController() + const result = queryEnvironment(process.execPath, ['-e', 'setInterval(() => {}, 1000)'], { ...options(), signal: lifetime.signal }) + lifetime.abort() + expect(await result).toEqual({ kind: 'missing', reason: 'query-failed' }) +}) diff --git a/packages/client/product-analytics/tests/environment.host.spec.ts b/packages/client/product-analytics/tests/environment.host.spec.ts new file mode 100644 index 0000000000..6fba26822d --- /dev/null +++ b/packages/client/product-analytics/tests/environment.host.spec.ts @@ -0,0 +1,179 @@ +import { afterEach, beforeEach, expect, it, vi } from 'vitest' +import * as os from 'node:os' +import { queryEnvironment } from '../src/environment-query.ts' +import { collectEnvironment } from '../src/environment.ts' + +vi.mock('node:os', async importOriginal => ({ ...await importOriginal(), + platform: vi.fn(), cpus: vi.fn(), totalmem: vi.fn(), release: vi.fn(), machine: vi.fn(), +})) +vi.mock('../src/environment-query.ts', () => ({ queryEnvironment: vi.fn() })) + +const options = () => ({ timeoutMillis: 5000, maxOutputBytes: 16384, signal: new AbortController().signal }) +const windows = { major: 10, minor: 0, build: '26300', revision: 9457 } +const processors = [{ cores: 8, architecture: 9 }, { cores: 6, architecture: 9 }] +const macVersion = 'ProductName: macOS\nProductVersion: 27.0.1\nBuildVersion: 26A434\n' +const macCpu = 'hw.physicalcpu: 8\nhw.optional.arm64: 1\n' + +beforeEach(() => { + vi.resetAllMocks() + vi.stubEnv('SystemRoot', 'C:\\Windows') + vi.mocked(os.platform).mockReturnValue('win32') + vi.mocked(os.release).mockReturnValue('10.0.26300') + vi.mocked(os.machine).mockReturnValue('x86_64') + vi.mocked(os.totalmem).mockReturnValue(32 * 1024 ** 3) + vi.mocked(os.cpus).mockReturnValue(Array.from({ length: 20 }, () => ({ + model: 'Fixture CPU', speed: 2400, times: { user: 0, nice: 0, sys: 0, idle: 0, irq: 0 }, + }))) +}) +afterEach(() => { vi.unstubAllEnvs() }) + +function outputs(version: unknown = windows, cpu: unknown = processors): void { + vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout: JSON.stringify(version) }) + .mockResolvedValueOnce({ kind: 'value', stdout: JSON.stringify(cpu) }) +} + +it('preserves Windows revision and sums physical cores across sockets independently of logical count', async () => { + outputs() + expect(await collectEnvironment(options())).toEqual({ + os_name: 'windows', os_version: 'win32 10.0.26300', os_version_full: '10.0.26300.9457', os_build: '26300', + os_arch: 'x64', app_arch: process.arch, cpu_model: 'Fixture CPU', cpu_logical_count: 20, + cpu_physical_core_count: 14, memory_total_bytes: 34359738368, missing_fields: {}, + }) + for (const [command, args] of vi.mocked(queryEnvironment).mock.calls) { + expect(command).toContain('powershell.exe') + expect(args.slice(0, 3)).toEqual(['-NoProfile', '-NonInteractive', '-EncodedCommand']) + const script = Buffer.from(args[3]!, 'base64').toString('utf16le') + expect(script).not.toMatch(/UUID|SerialNumber|ProductId|Hostname|Win32_ComputerSystemProduct/i) + } +}) + +it.each([0, 9457, 9458])('retains a genuine Windows revision of %s', async (revision) => { + outputs({ ...windows, revision }) + expect((await collectEnvironment(options())).os_version_full).toBe(`10.0.26300.${revision}`) +}) + +it('does not invent revision zero when UBR is missing', async () => { + outputs({ ...windows, revision: null }) + const result = await collectEnvironment(options()) + expect(result.os_build).toBe('26300') + expect(result).not.toHaveProperty('os_version_full') + expect(result.missing_fields).toEqual({ os_version_full: 'unavailable' }) +}) + +it('marks missing build, major, and minor without discarding CPU metadata', async () => { + outputs({ major: null, minor: null, build: null, revision: null }) + const result = await collectEnvironment(options()) + expect(result.missing_fields).toEqual({ os_version_full: 'unavailable', os_build: 'unavailable' }) + expect(result.cpu_physical_core_count).toBe(14) +}) + +it.each([ + { major: -1, minor: 0, build: '26300', revision: 9457 }, + { ...windows, revision: '9457' }, { ...windows, build: '26300.9457' }, +])('rejects invalid Windows version JSON: %j', async (version) => { + outputs(version) + expect((await collectEnvironment(options())).missing_fields).toMatchObject({ os_version_full: 'invalid-value', os_build: 'invalid-value' }) +}) + +it.each([{ cpu: [] }, { cpu: [{ cores: 0, architecture: 9 }] }, { cpu: [{ cores: '8', architecture: 9 }] }])('rejects invalid CPU fields: %j', async ({ cpu }) => { + outputs(windows, cpu) + expect((await collectEnvironment(options())).missing_fields).toEqual({ os_arch: 'invalid-value', cpu_physical_core_count: 'invalid-value' }) +}) + +it.each([null, 99])('marks unavailable CPU architecture %s and physical cores', async (architecture) => { + outputs(windows, [{ cores: null, architecture }]) + const result = await collectEnvironment(options()) + expect(result.missing_fields).toEqual({ os_arch: 'unavailable', cpu_physical_core_count: 'unavailable' }) + expect(result).not.toHaveProperty('cpu_physical_core_count') +}) + +it('does not merge conflicting socket architectures', async () => { + outputs(windows, [{ cores: 4, architecture: 9 }, { cores: 4, architecture: 12 }]) + expect((await collectEnvironment(options())).missing_fields.os_arch).toBe('unavailable') +}) + +it('reads native Windows ARM64 independently of the application binary', async () => { + outputs(windows, [{ cores: 8, architecture: 12 }]) + const result = await collectEnvironment(options()) + expect(result.os_arch).toBe('arm64') + expect(result.app_arch).toBe(process.arch) +}) + +it.each([undefined, ''])('omits Windows query fields when SystemRoot is %s', async (root) => { + vi.stubEnv('SystemRoot', root) + const result = await collectEnvironment(options()) + expect(queryEnvironment).not.toHaveBeenCalled() + expect(result.missing_fields).toEqual({ os_version_full: 'unavailable', os_build: 'unavailable', os_arch: 'unavailable', cpu_physical_core_count: 'unavailable' }) +}) + +it.each(['timeout', 'output-limit', 'query-failed'] as const)('isolates %s from local CPU and memory fields', async (reason) => { + vi.mocked(queryEnvironment).mockResolvedValue({ kind: 'missing', reason }) + const result = await collectEnvironment(options()) + expect(result.cpu_model).toBe('Fixture CPU') + expect(result.memory_total_bytes).toBe(34359738368) + expect(result.missing_fields).toEqual({ os_version_full: reason, os_build: reason, os_arch: reason, cpu_physical_core_count: reason }) +}) + +it('rejects malformed JSON without exposing raw output', async () => { + vi.mocked(queryEnvironment).mockResolvedValue({ kind: 'value', stdout: 'private output' }) + const result = await collectEnvironment(options()) + expect(result.missing_fields.os_version_full).toBe('invalid-value') + expect(JSON.stringify(result)).not.toContain('private') +}) + +it.each(['1', '0'])('reports macOS product version and native architecture when arm64=%s', async (arm) => { + vi.mocked(os.platform).mockReturnValue('darwin') + vi.mocked(os.release).mockReturnValue('26.0.0') + vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout: macVersion }) + .mockResolvedValueOnce({ kind: 'value', stdout: macCpu.replace('arm64: 1', `arm64: ${arm}`) }) + const result = await collectEnvironment(options()) + expect(result).toMatchObject({ os_name: 'macos', os_version: 'darwin 26.0.0', os_version_full: '27.0.1', os_build: '26A434', + os_arch: arm === '1' ? 'arm64' : 'x64', app_arch: process.arch, cpu_physical_core_count: 8, missing_fields: {} }) + expect(vi.mocked(queryEnvironment).mock.calls.map(([command, args]) => [command, args])).toEqual([ + ['/usr/bin/sw_vers', []], ['/usr/sbin/sysctl', ['-i', 'hw.physicalcpu', 'hw.optional.arm64']], + ]) +}) + +it('preserves macOS security response suffixes', async () => { + vi.mocked(os.platform).mockReturnValue('darwin') + vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout: `${macVersion}ProductVersionExtra: (a)\n` }) + .mockResolvedValueOnce({ kind: 'value', stdout: macCpu }) + expect((await collectEnvironment(options())).os_version_full).toBe('27.0.1 (a)') +}) + +it.each(['', 'ProductVersion: garbage\nBuildVersion: private\nProductVersionExtra: invalid', + 'ProductVersion: 27.0.1\nBuildVersion: bad\nProductVersionExtra: invalid'])('rejects invalid macOS metadata: %s', async (stdout) => { + vi.mocked(os.platform).mockReturnValue('darwin') + vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout }) + .mockResolvedValueOnce({ kind: 'value', stdout: 'hw.physicalcpu: 0\nhw.optional.arm64: unknown' }) + expect((await collectEnvironment(options())).missing_fields).toEqual({ + os_version_full: 'invalid-value', os_build: 'invalid-value', cpu_physical_core_count: 'invalid-value', os_arch: 'unavailable', + }) +}) + +it('reports macOS query failure independently of Node fields', async () => { + vi.mocked(os.platform).mockReturnValue('darwin') + vi.mocked(queryEnvironment).mockResolvedValue({ kind: 'missing', reason: 'timeout' }) + expect((await collectEnvironment(options())).missing_fields).toEqual({ os_version_full: 'timeout', os_build: 'timeout', os_arch: 'timeout', cpu_physical_core_count: 'timeout' }) +}) + +it.each([0, Number.NaN, 1.5])('marks unavailable memory %s instead of reporting zero', async (memory) => { + vi.mocked(os.totalmem).mockReturnValue(memory) + vi.mocked(os.cpus).mockReturnValue([]) + outputs() + const result = await collectEnvironment(options()) + expect(result).not.toHaveProperty('memory_total_bytes') + expect(result.missing_fields).toMatchObject({ memory_total_bytes: 'unavailable', cpu_model: 'unavailable', cpu_logical_count: 'unavailable' }) +}) + +it('isolates refused synchronous OS sources', async () => { + for (const source of [os.totalmem, os.cpus, os.release]) vi.mocked(source).mockImplementation(() => { throw new Error('source refused') }) + outputs() + expect((await collectEnvironment(options())).missing_fields).toMatchObject({ os_version: 'unavailable', memory_total_bytes: 'unavailable', cpu_model: 'unavailable', cpu_logical_count: 'unavailable' }) +}) + +it('marks unsupported platform queries without spawning commands', async () => { + vi.mocked(os.platform).mockReturnValue('linux') + expect((await collectEnvironment(options())).missing_fields).toEqual({ os_version_full: 'unsupported', os_build: 'unsupported', cpu_physical_core_count: 'unsupported' }) + expect(queryEnvironment).not.toHaveBeenCalled() +}) diff --git a/packages/client/product-analytics/tests/fixtures/environment.host.ts b/packages/client/product-analytics/tests/fixtures/environment.host.ts new file mode 100644 index 0000000000..a19cd60755 --- /dev/null +++ b/packages/client/product-analytics/tests/fixtures/environment.host.ts @@ -0,0 +1,66 @@ +/** Real Host environment collection through the shipped profile and isolated OTLP receiver. */ +import { createServer } from 'node:http' +import { once } from 'node:events' +import { writeFile } from 'node:fs/promises' +import { execFile } from 'node:child_process' +import { totalmem, cpus, platform } from 'node:os' +import { promisify } from 'node:util' +import { join } from 'node:path' +import type {} from '@deepseek-ai/dsh-client-product-analytics' +import { bootProductionProfile } from '../../../../test-support/loader-smoke/tests/fixtures/production-profile.ts' + +const overlay = process.argv[2] +if (overlay === undefined) throw new Error('environment fixture requires its overlay') +const captures: unknown[] = [] +const receiver = createServer((req, res) => { + const chunks: Buffer[] = [] + req.on('data', (chunk: Buffer) => { chunks.push(chunk) }) + req.on('end', () => { captures.push(JSON.parse(Buffer.concat(chunks).toString())); res.end('{}') }) +}) +const run = promisify(execFile) +const reference: Record = { memory_total_bytes: totalmem(), cpu_logical_count: cpus().length } +try { + receiver.listen(0, '127.0.0.1') + await once(receiver, 'listening') + const address = receiver.address() + if (address === null || typeof address === 'string') throw new Error('environment collector has no port') + process.env.DSH_PRODUCT_TELEMETRY_TEST_ENDPOINT = `http://127.0.0.1:${address.port}/logs` + const ctx = await bootProductionProfile({ binName: 'environment-test', profile: 'headless', overlayPaths: [overlay], + prepare: (root) => { root.provide('deepseekAccount', { getDeviceIdentity: async () => undefined } as never) }, + }) + try { + const analytics = ctx.get('productAnalytics') + if (analytics === undefined) throw new Error('environment analytics did not mount') + await analytics.report({ eventName: 'desktop_app_launch', timestamp: 1_800_000_000_000, attributes: {} }) + } finally { await ctx.fiber.dispose() } + if (platform() === 'darwin') { + reference.os_version_full = (await run('/usr/bin/sw_vers', ['-productVersion'])).stdout.trim() + reference.os_build = (await run('/usr/bin/sw_vers', ['-buildVersion'])).stdout.trim() + reference.cpu_physical_core_count = Number((await run('/usr/sbin/sysctl', ['-n', 'hw.physicalcpu'])).stdout.trim()) + reference.os_arch = (await run('/usr/sbin/sysctl', ['-n', 'hw.optional.arm64'])).stdout.trim() === '1' ? 'arm64' : 'x64' + } else if (platform() === 'win32') { + const systemRoot = process.env.SystemRoot + if (systemRoot === undefined) throw new Error('Windows fixture requires SystemRoot') + const script = ` +$r = Get-ItemProperty 'HKLM:\\SOFTWARE\\Microsoft\\Windows NT\\CurrentVersion' -Name CurrentMajorVersionNumber,CurrentMinorVersionNumber,CurrentBuildNumber,UBR +Write-Output "$($r.CurrentMajorVersionNumber).$($r.CurrentMinorVersionNumber).$($r.CurrentBuildNumber).$($r.UBR)" +Write-Output $r.CurrentBuildNumber +Write-Output ((Get-CimInstance Win32_Processor | Measure-Object NumberOfCores -Sum).Sum) +Write-Output (Get-CimInstance Win32_Processor | Select-Object -First 1 -ExpandProperty Architecture) +` + const { stdout } = await run(join(systemRoot, 'System32', 'WindowsPowerShell', 'v1.0', 'powershell.exe'), + ['-NoProfile', '-NonInteractive', '-EncodedCommand', Buffer.from(script, 'utf16le').toString('base64')]) + const [version, build, cores, architecture] = stdout.trim().split(/\r?\n/) + if (version === undefined || build === undefined) throw new Error('Windows fixture returned no version') + reference.os_version_full = version + reference.os_build = build + reference.cpu_physical_core_count = Number(cores) + reference.os_arch = ({ 0: 'ia32', 5: 'arm', 9: 'x64', 12: 'arm64' } as Record)[Number(architecture)]! + } + await writeFile('captures.json', JSON.stringify({ captures, reference })) +} finally { + const closed = once(receiver, 'close') + receiver.close() + receiver.closeAllConnections() + await closed +} diff --git a/packages/client/product-analytics/tests/fixtures/environment.patch.yml b/packages/client/product-analytics/tests/fixtures/environment.patch.yml new file mode 100644 index 0000000000..8d150bedc1 --- /dev/null +++ b/packages/client/product-analytics/tests/fixtures/environment.patch.yml @@ -0,0 +1,19 @@ +- id: session-telemetry-otel + disabled: true +- id: headless-startup + disabled: true +- id: headless-runner + disabled: true +- insert: + - id: product-telemetry-otel + name: '@deepseek-ai/dsh-host-product-telemetry-otel' + config: + endpoint: !!js process.env.DSH_PRODUCT_TELEMETRY_TEST_ENDPOINT + serviceName: dsh-environment-test + serviceVersion: synthetic-release + compression: none + - id: product-analytics + name: '@deepseek-ai/dsh-client-product-analytics' + config: + enabled: true + appVersion: synthetic-release diff --git a/packages/client/product-analytics/tsconfig.host.json b/packages/client/product-analytics/tsconfig.host.json index 487bb221ac..b4b4e831ab 100644 --- a/packages/client/product-analytics/tsconfig.host.json +++ b/packages/client/product-analytics/tsconfig.host.json @@ -7,7 +7,9 @@ }, "files": [ "src/index.ts", - "src/events.ts" + "src/events.ts", + "src/environment.ts", + "src/environment-query.ts" ], "references": [ { diff --git a/packages/extensions/tool-cordis/src/api-catalog.ts b/packages/extensions/tool-cordis/src/api-catalog.ts index 798e310219..708eee6876 100644 --- a/packages/extensions/tool-cordis/src/api-catalog.ts +++ b/packages/extensions/tool-cordis/src/api-catalog.ts @@ -1797,7 +1797,7 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [ }, { signature: '@Remote async report(event: ProductEvent): Promise', - description: 'Submit selected Desktop fields; missing identity is omitted and never generated.', + description: 'Submit selected Desktop fields with cached static environment and fresh identity. The first intake waits for bounded environment queries; concurrent intakes share them. Missing identity is omitted; failed environment fields carry fixed reasons.', parameters: [{ name: 'event', description: 'typed product event without message contents or credentials.' }], returns: 'after local submission; no delivery or warehouse acknowledgement.', }, diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml index 977eeaa9a8..ed7aa7dd8b 100644 --- a/pnpm-lock.yaml +++ b/pnpm-lock.yaml @@ -3175,6 +3175,9 @@ importers: '@deepseek-ai/dsh-api-gateway': specifier: workspace:* version: link:../../api/gateway + '@deepseek-ai/dsh-app-boot': + specifier: workspace:* + version: link:../../boot/app-boot '@deepseek-ai/dsh-compaction': specifier: workspace:* version: link:../../compaction/compaction @@ -3190,6 +3193,9 @@ importers: '@deepseek-ai/dsh-llm': specifier: workspace:* version: link:../../llm/llm + '@deepseek-ai/dsh-loader-smoke': + specifier: workspace:* + version: link:../../test-support/loader-smoke '@deepseek-ai/dsh-otel': specifier: workspace:* version: link:../../telemetry/otel diff --git a/tsconfig.client.json b/tsconfig.client.json index fc87ea5394..5d7035ed6f 100644 --- a/tsconfig.client.json +++ b/tsconfig.client.json @@ -48,7 +48,9 @@ // belongs to the host aggregate, which excludes this program's `*.client.*` // in turn. `exclude` wins over `include`, so the test glob above stays broad. "exclude": [ - "packages/client/*/tests/**/*.host.spec.ts" + "packages/client/*/tests/**/*.host.spec.ts", + "packages/client/*/tests/**/*.host.e2e.ts", + "packages/client/*/tests/**/*.host.ts" ], "references": [ { "path": "./packages/experimental/client-ui-cot-translation/tsconfig.client.json" }, From 04c94890658c3f87a93deb9d6396402d44755818 Mon Sep 17 00:00:00 2001 From: 07akioni <07akioni2@gmail.com> Date: Wed, 14 Oct 2026 16:10:40 +0800 Subject: [PATCH 025/162] fix(telemetry): handle native metadata edge cases and scalar failures --- docs/subsystems/product-telemetry.i18n.yaml | 4 +-- docs/subsystems/product-telemetry.md | 2 +- docs/subsystems/product-telemetry.zh.md | 2 +- .../client/product-analytics/README.i18n.yaml | 8 ++--- packages/client/product-analytics/README.md | 10 ++++--- .../client/product-analytics/README.zh.md | 10 ++++--- .../src/environment-query.ts | 26 +++++++++------- .../product-analytics/src/environment.ts | 20 +++++++------ .../client/product-analytics/src/index.ts | 6 +++- .../tests/collection.host.spec.ts | 25 +++++++++++++--- .../tests/environment-composition.host.e2e.ts | 5 ++-- .../tests/environment-query.host.spec.ts | 30 ++++++++++++++++++- .../tests/environment.host.spec.ts | 23 +++++++++++--- .../tests/fixtures/environment.host.ts | 8 +++-- tsconfig.client.json | 4 +-- 15 files changed, 131 insertions(+), 52 deletions(-) diff --git a/docs/subsystems/product-telemetry.i18n.yaml b/docs/subsystems/product-telemetry.i18n.yaml index 28b4e64b14..0ad595e309 100644 --- a/docs/subsystems/product-telemetry.i18n.yaml +++ b/docs/subsystems/product-telemetry.i18n.yaml @@ -3,8 +3,8 @@ # After editing either side, bring the other along and re-record with: # pnpm run verify-translation-pairing --write docs/subsystems/product-telemetry.md /product-telemetry: - en: 04e9a32dfbe69355 - zh: 0e6c78274dd97466 + en: 419a470ec1cce036 + zh: 91e69fb502318e45 /product-telemetry/record-types: en: bf5eadedbc896f5a zh: 158fce1609a85d6d diff --git a/docs/subsystems/product-telemetry.md b/docs/subsystems/product-telemetry.md index 9d656631f6..14cf59985b 100644 --- a/docs/subsystems/product-telemetry.md +++ b/docs/subsystems/product-telemetry.md @@ -2,7 +2,7 @@ English | [中文](product-telemetry.zh.md) -The [product telemetry plugin](../../packages/host/product-telemetry-otel/README.md) sends explicitly selected analytics events through OTLP/HTTP. Its `productTelemetry` service owns submission only; product consumers own when an event occurs and which fields are approved. The exporter collects no Session data or identifiers automatically. The [Desktop analytics consumer](../../packages/client/product-analytics/README.md) selects interaction and live compaction events, adds available login identity, and follows the Desktop launch-time collection switch. Ordinary Web clients do not collect product events. +The [product telemetry plugin](../../packages/host/product-telemetry-otel/README.md) sends explicitly selected analytics events through OTLP/HTTP. Its `productTelemetry` service owns submission only; product consumers own when an event occurs and which fields are approved. The exporter collects no Session data or identifiers automatically. The [Desktop analytics consumer](../../packages/client/product-analytics/README.md) selects interaction and live compaction events, adds cached static environment and available login identity, and follows the Desktop launch-time collection switch. Ordinary Web clients do not collect product events. `ProductTelemetryRecord` requires an event name, string body, and Unix-millisecond timestamp. Attributes accept scalar strings, numbers, booleans, and one-level objects of those values (`ProductTelemetryScalar`). Optional severity uses OTel severity numbers and defaults to INFO. Observation time is assigned when the record is enqueued. diff --git a/docs/subsystems/product-telemetry.zh.md b/docs/subsystems/product-telemetry.zh.md index e8072807ec..feffdfae91 100644 --- a/docs/subsystems/product-telemetry.zh.md +++ b/docs/subsystems/product-telemetry.zh.md @@ -2,7 +2,7 @@ [English](product-telemetry.md) | 中文 -[产品埋点插件](../../packages/host/product-telemetry-otel/README.zh.md) 通过 OTLP/HTTP 发送明确选定的分析事件。`productTelemetry` 服务仅负责提交；产品消费方决定事件发生时机与获准采集的字段。导出器不自动采集 Session 数据或标识。[桌面埋点消费方](../../packages/client/product-analytics/README.zh.md)选择交互与实时压缩事件，补充可用的登录身份，并遵守桌面启动时的采集开关。普通 Web 客户端不采集产品事件。 +[产品埋点插件](../../packages/host/product-telemetry-otel/README.zh.md) 通过 OTLP/HTTP 发送明确选定的分析事件。`productTelemetry` 服务仅负责提交；产品消费方决定事件发生时机与获准采集的字段。导出器不自动采集 Session 数据或标识。[桌面埋点消费方](../../packages/client/product-analytics/README.zh.md)选择交互与实时压缩事件，补充缓存的静态环境和可用的登录身份，并遵守桌面启动时的采集开关。普通 Web 客户端不采集产品事件。 `ProductTelemetryRecord` 要求事件名称、字符串 body 和 Unix 毫秒时间戳。属性接受字符串、数字、布尔标量，以及这些值组成的单层对象（`ProductTelemetryScalar`）。可选的严重程度使用 OTel 严重程度数字，默认为 INFO。记录入队时填写观测时间。 diff --git a/packages/client/product-analytics/README.i18n.yaml b/packages/client/product-analytics/README.i18n.yaml index 989c03c5b2..546332a531 100644 --- a/packages/client/product-analytics/README.i18n.yaml +++ b/packages/client/product-analytics/README.i18n.yaml @@ -15,8 +15,8 @@ en: 2fee9973f19f9401 zh: 8ab4b55413cb8408 /deepseek-ai-dsh-client-product-analytics/use-this-package: - en: 062e1847cfcc60ee - zh: ea1b6cc2d2413226 + en: 042e8eef8647b025 + zh: 256ff24fee965eda /deepseek-ai-dsh-client-product-analytics/understand-the-implementation: en: 96b1169d78f4c761 zh: 475df4ea6ed1cb06 @@ -27,8 +27,8 @@ en: 1f5125831808e7a9 zh: 97566fdde332df39 /deepseek-ai-dsh-client-product-analytics/known-limitations-and-deferred-work: - en: a33ad31745fd5240 - zh: 8b579cfc4c9b8155 + en: 47bd8bf635e5edbb + zh: defee1a5d6764bfe /deepseek-ai-dsh-client-product-analytics/known-limitations-and-deferred-work/dev-note: en: 0a1253c6f7b5e8b9 zh: 9f56528562c7c914 diff --git a/packages/client/product-analytics/README.md b/packages/client/product-analytics/README.md index 1317bcd5fe..54cd8a820f 100644 --- a/packages/client/product-analytics/README.md +++ b/packages/client/product-analytics/README.md @@ -29,19 +29,19 @@ Every product event carries available `device_id`, `user_id`, and `app_version` | Field | Meaning and source | |---|---| -| `os_name` | `windows` or `macos`; Node platform selects the mapping | +| `os_name` | `windows` or `macos`; other platforms retain the raw Node platform name (such as `linux`) | | `os_version` | Legacy ` ` from `node:os.release()` | | `os_version_full` | Windows registry major/minor/build plus `UBR` revision; macOS `sw_vers` product version including a security-response suffix when present | | `os_build` | Windows registry `CurrentBuildNumber`; macOS `sw_vers` build | -| `os_arch` | Native architecture mapped to `ia32`, `x64`, `arm`, or `arm64`; Windows processor CIM architecture or macOS `hw.optional.arm64`, independent of application emulation | +| `os_arch` | Native architecture mapped to `ia32`, `x64`, `arm`, or `arm64`; Windows processor CIM architecture or macOS `hw.optional.arm64` (`1` means ARM64; an absent key means Intel x64), independent of application emulation | | `app_arch` | Running binary architecture from `process.arch` | | `cpu_model` | First logical processor's model from `node:os.cpus()` | | `cpu_logical_count` | Number of system-reported logical processors from `node:os.cpus()`, not application parallelism | | `cpu_physical_core_count` | Sum of Windows processor CIM `NumberOfCores` across sockets; macOS `hw.physicalcpu` | | `memory_total_bytes` | Total physical memory from `node:os.totalmem()`, in bytes without rounding | -| `missing_fields` | Object mapping omitted environment field names to `unavailable`, `invalid-value`, `query-failed`, `timeout`, `output-limit`, or `unsupported`; empty when all fields are available | +| `missing_fields` | Sorted comma-separated `field=reason` pairs using `unavailable`, `invalid-value`, `query-failed`, `timeout`, `output-limit`, or `unsupported`; omitted when all fields are available | -The first enabled intake collects one environment snapshot; concurrent and later intakes reuse its successful fields and failures until the plugin remounts or the Host restarts. Intake waits for sampling while callers submit asynchronously; startup does not await collection. Identity changes remain visible on subsequent events. Each metadata command has a configurable `environmentQueryTimeoutMillis` deadline (5,000 ms) and `environmentQueryMaxOutputBytes` bound per output stream (16,384 bytes). Disposal cancels commands and waits for child closure; disabled or disposed intake never emits late results. Missing revision never becomes `.0`, and missing numeric fields never become zero. Only Windows and macOS have full-version and physical-core queries; other platforms mark those fields `unsupported`. +The first enabled intake collects one environment snapshot; concurrent and later intakes reuse its successful fields and failures until the plugin remounts or the Host restarts. Intake waits for sampling while callers submit asynchronously; startup does not await collection. Identity changes remain visible on subsequent events. Each metadata command has a configurable `environmentQueryTimeoutMillis` deadline (5,000 ms) and `environmentQueryMaxOutputBytes` bound per output stream (16,384 bytes). Disposal cancels commands and waits for child closure; disabled or disposed intake never emits late results. Missing revision never becomes `.0`, and missing numeric fields never become zero. Only Windows and macOS have full-version, native-architecture, and physical-core queries; other platforms mark those fields `unsupported`. Electron forwards its build-inlined `DSH_CLIENT_VERSION` to the Host; analytics and the exporter reuse this client version. The exporter requires it. The [Desktop composition](../../bundle/web-app/README.md) owns batching and timeout settings, including cancellation at the shutdown deadline. @@ -74,6 +74,8 @@ None; collection does not modify model requests. Delivery is best effort. +- Environment sampling is one-shot: startup contention can cause a cached timeout for the whole run. `missing_fields` describes sampling failure, not hardware absence; a later event does not retry. Deployments can increase the query deadline; remount or restart samples again. +- Host security policies can block or terminate the fixed PowerShell queries, producing `query-failed` on managed Windows machines. - Disabling stops new collection; it does not clear the exporter queue. - Renderer failures are discarded, and the exporter has no durable outbox or warehouse acknowledgement. - Native startup reporting waits for Host authentication; a process that fails before Host readiness cannot report its launch. diff --git a/packages/client/product-analytics/README.zh.md b/packages/client/product-analytics/README.zh.md index c5d588b851..803494c0a5 100644 --- a/packages/client/product-analytics/README.zh.md +++ b/packages/client/product-analytics/README.zh.md @@ -29,19 +29,19 @@ kind: "package-reference" | 字段 | 含义与来源 | |---|---| -| `os_name` | `windows` 或 `macos`；由 Node 平台选择映射 | +| `os_name` | `windows` 或 `macos`；其他平台保留原始 Node 平台名称（如 `linux`） | | `os_version` | 原有的 ` <内核版本>`，来自 `node:os.release()` | | `os_version_full` | Windows 注册表 major/minor/build 加上 `UBR` 修订号；macOS `sw_vers` 产品版本，包括存在时的安全响应后缀 | | `os_build` | Windows 注册表 `CurrentBuildNumber`；macOS `sw_vers` build | -| `os_arch` | 原生架构映射为 `ia32`、`x64`、`arm` 或 `arm64`；来自 Windows 处理器 CIM 架构或 macOS `hw.optional.arm64`，独立于应用仿真运行 | +| `os_arch` | 原生架构映射为 `ia32`、`x64`、`arm` 或 `arm64`；来自 Windows 处理器 CIM 架构或 macOS `hw.optional.arm64`（`1` 表示 ARM64，缺少键表示 Intel x64），独立于应用仿真运行 | | `app_arch` | 来自 `process.arch` 的运行二进制架构 | | `cpu_model` | 来自 `node:os.cpus()` 的首个逻辑处理器型号 | | `cpu_logical_count` | 来自 `node:os.cpus()` 的系统报告逻辑处理器数量，不表示应用可用并行度 | | `cpu_physical_core_count` | Windows 各 socket 处理器 CIM `NumberOfCores` 之和；macOS `hw.physicalcpu` | | `memory_total_bytes` | 来自 `node:os.totalmem()` 的总物理内存，单位 bytes，不取整 | -| `missing_fields` | 对象，将省略的环境字段名映射到 `unavailable`、`invalid-value`、`query-failed`、`timeout`、`output-limit` 或 `unsupported`；全部字段可用时为空 | +| `missing_fields` | 按字段名排序、逗号分隔的 `field=reason` 项，原因使用 `unavailable`、`invalid-value`、`query-failed`、`timeout`、`output-limit` 或 `unsupported`；全部字段可用时省略 | -首次启用状态下的事件接收采集一份环境快照；并发和后续事件复用成功字段及失败状态，直到插件重新挂载或 Host 重启。事件接收等待采样，调用方异步提交；启动不等待采集。后续事件仍能反映身份变化。每条元数据命令具有可配置的 `environmentQueryTimeoutMillis` 期限（5,000 毫秒）和每路输出的 `environmentQueryMaxOutputBytes` 上限（16,384 bytes）。卸载取消命令并等待子进程关闭；禁用或卸载后的事件接收不会发送迟到结果。缺失修订号不会变为 `.0`，缺失数字字段不会变为零。只有 Windows 和 macOS 查询完整版本与物理核心数；其他平台将这些字段标记为 `unsupported`。 +首次启用状态下的事件接收采集一份环境快照；并发和后续事件复用成功字段及失败状态，直到插件重新挂载或 Host 重启。事件接收等待采样，调用方异步提交；启动不等待采集。后续事件仍能反映身份变化。每条元数据命令具有可配置的 `environmentQueryTimeoutMillis` 期限（5,000 毫秒）和每路输出的 `environmentQueryMaxOutputBytes` 上限（16,384 bytes）。卸载取消命令并等待子进程关闭；禁用或卸载后的事件接收不会发送迟到结果。缺失修订号不会变为 `.0`，缺失数字字段不会变为零。只有 Windows 和 macOS 查询完整版本、原生架构与物理核心数；其他平台将这些字段标记为 `unsupported`。 Electron 将构建内联的 `DSH_CLIENT_VERSION` 传给 Host；埋点与导出器复用这一客户端版本，导出器要求该值存在。[桌面组合](../../bundle/web-app/README.zh.md)负责批量发送与超时配置，包括达到关闭期限时的取消行为。 @@ -74,6 +74,8 @@ Electron 将构建内联的 `DSH_CLIENT_VERSION` 传给 Host；埋点与导出 交付采用尽力而为的方式。 +- 环境仅采样一次：启动期间的资源争用可能导致超时，并缓存到本次运行结束。`missing_fields` 表示采样失败，不表示硬件缺失；后续事件不重试。部署可提高查询期限；重新挂载或重启会再次采样。 +- 主机安全策略可能阻止或终止固定的 PowerShell 查询，使受管理的 Windows 机器产生 `query-failed`。 - 关闭只停止新采集，不清空导出器队列。 - 渲染进程的发送失败会被丢弃，导出器没有持久化待发队列或数据仓库确认。 - 原生启动事件等待 Host 认证完成；若进程在 Host 就绪前失败，则无法上报启动事件。 diff --git a/packages/client/product-analytics/src/environment-query.ts b/packages/client/product-analytics/src/environment-query.ts index 8086fff82f..ae0eb79551 100644 --- a/packages/client/product-analytics/src/environment-query.ts +++ b/packages/client/product-analytics/src/environment-query.ts @@ -1,5 +1,5 @@ /** Bounded, cancellable queries of fixed operating-system metadata commands. */ -import { execFile } from 'node:child_process' +import { execFile, type ChildProcess } from 'node:child_process' /** Fixed reasons exposed instead of command output or raw errors. */ export type MissingReason = 'unavailable' | 'invalid-value' | 'query-failed' | 'timeout' | 'output-limit' | 'unsupported' @@ -26,15 +26,21 @@ export interface QueryOptions { */ export async function queryEnvironment(command: string, args: readonly string[], options: QueryOptions): Promise { const result = Promise.withResolvers() - const child = execFile(command, [...args], { - encoding: 'utf8', windowsHide: true, signal: options.signal, killSignal: 'SIGKILL', - timeout: options.timeoutMillis, maxBuffer: options.maxOutputBytes, - }, (error, stdout) => { - if (error === null) { result.resolve({ kind: 'value', stdout }); return } - const reason = error.code === 'ERR_CHILD_PROCESS_STDIO_MAXBUFFER' ? 'output-limit' - : error.killed && !options.signal.aborted ? 'timeout' : 'query-failed' - result.resolve({ kind: 'missing', reason }) - }) + let child: ChildProcess + try { + child = execFile(command, [...args], { + encoding: 'utf8', windowsHide: true, signal: options.signal, killSignal: 'SIGKILL', + timeout: options.timeoutMillis, maxBuffer: options.maxOutputBytes, + }, (error, stdout) => { + if (error === null) { result.resolve({ kind: 'value', stdout }); return } + const reason = error.code === 'ERR_CHILD_PROCESS_STDIO_MAXBUFFER' ? 'output-limit' + : error.killed && !options.signal.aborted ? 'timeout' : 'query-failed' + result.resolve({ kind: 'missing', reason }) + }) + } catch (_error) { + // Synchronous OS spawn failures create no child; other queries remain owned. + return { kind: 'missing', reason: 'query-failed' } + } // execFile can report abort before stdio closes; disposal owns both settlements. const closed = new Promise((resolve) => { child.once('close', () => { resolve() }) }) const [output] = await Promise.all([result.promise, closed]) diff --git a/packages/client/product-analytics/src/environment.ts b/packages/client/product-analytics/src/environment.ts index 7f067cc7b8..60ecd91b9e 100644 --- a/packages/client/product-analytics/src/environment.ts +++ b/packages/client/product-analytics/src/environment.ts @@ -1,5 +1,5 @@ /** Static Desktop environment fields, independent of account and credential availability. */ -import { cpus, machine, platform, release, totalmem } from 'node:os' +import { cpus, platform, release, totalmem } from 'node:os' import { join } from 'node:path' import { z } from 'zod' import { queryEnvironment, type MissingReason, type QueryOptions, type QueryResult } from './environment-query.ts' @@ -27,7 +27,7 @@ export interface EnvironmentAttributes { /** Total physical memory in bytes. */ memory_total_bytes?: number /** Only failed fields, with no raw commands, output, or errors. */ - missing_fields: Record + missing_fields: Partial> } type EnvironmentField = Exclude @@ -126,9 +126,12 @@ export async function collectEnvironment(options: QueryOptions): Promise machine()) - missing(['os_version_full', 'os_build', 'cpu_physical_core_count'], 'unsupported') + missing(['os_version_full', 'os_build', 'os_arch', 'cpu_physical_core_count'], 'unsupported') } return fields } diff --git a/packages/client/product-analytics/src/index.ts b/packages/client/product-analytics/src/index.ts index ce497f99df..86a08ef60a 100644 --- a/packages/client/product-analytics/src/index.ts +++ b/packages/client/product-analytics/src/index.ts @@ -109,11 +109,15 @@ export default class ProductAnalytics extends TypertRemoteService { this.environment, this.ctx.deepseekAccount.getDeviceIdentity().catch(() => undefined), ]) if (!this.enabled()) return + const { missing_fields: missingFields, ...metadata } = environment + const missing = Object.entries(missingFields).sort(([a], [b]) => a.localeCompare(b)) + .map(([field, reason]) => `${field}=${reason}`).join(',') this.ctx.productTelemetry.emit({ ...event, body: event.eventName, attributes: { ...event.attributes, - ...environment, + ...metadata, + ...missing === '' ? {} : { missing_fields: missing }, ...identity?.deviceId === undefined ? {} : { device_id: identity.deviceId }, ...identity?.userId === undefined ? {} : { user_id: identity.userId }, ...this.config.appVersion === undefined ? {} : { app_version: this.config.appVersion }, diff --git a/packages/client/product-analytics/tests/collection.host.spec.ts b/packages/client/product-analytics/tests/collection.host.spec.ts index 5538f4c513..1ccd331183 100644 --- a/packages/client/product-analytics/tests/collection.host.spec.ts +++ b/packages/client/product-analytics/tests/collection.host.spec.ts @@ -15,6 +15,7 @@ const environment: EnvironmentAttributes = { os_arch: 'arm64', app_arch: 'x64', cpu_model: 'Fixture CPU', cpu_logical_count: 16, cpu_physical_core_count: 8, memory_total_bytes: 34359738368, missing_fields: {}, } +const { missing_fields: _missingFields, ...commonEnvironment } = environment beforeEach(() => { vi.resetAllMocks(); vi.mocked(collectEnvironment).mockResolvedValue(environment) }) const cleanup: (() => Promise)[] = [] @@ -47,16 +48,17 @@ it('reuses login identity and copies only approved common fields', async () => { expect(b.getDeviceIdentity).toHaveBeenCalledWith() expect(b.emit).toHaveBeenCalledExactlyOnceWith({ eventName: 'auth_page_click', body: 'auth_page_click', timestamp: 100, - attributes: { ...environment, button_name: 'sign_in', device_id: 'login-device', user_id: 'user-1', app_version: 'test-version' }, + attributes: { ...commonEnvironment, button_name: 'sign_in', device_id: 'login-device', user_id: 'user-1', app_version: 'test-version' }, }) expect(JSON.stringify(b.emit.mock.calls)).not.toContain('private-') + expect(b.emit.mock.calls[0]![0].attributes).not.toHaveProperty('missing_fields') }) it('missing identity does not discard an otherwise valid event', async () => { const b = await setup(true) b.getDeviceIdentity.mockRejectedValueOnce(new Error('credential store unavailable')) await b.ctx.productAnalytics.report({ eventName: 'plugin_add_button_click', timestamp: 100, attributes: {} }) - expect(b.emit.mock.calls[0]?.[0].attributes).toEqual({ ...environment, app_version: 'test-version' }) + expect(b.emit.mock.calls[0]?.[0].attributes).toEqual({ ...commonEnvironment, app_version: 'test-version' }) }) it('unload suppresses an identity lookup that settles after disposal', async () => { @@ -92,7 +94,10 @@ it('writes the selected event through the real exporter to an isolated collector const analytics = await ctx.plugin(Analytics, { enabled: true }) await ctx.productAnalytics.report({ eventName: 'api_key_save_click', timestamp: 1_800_000_000_000, attributes: {} }) await analytics.dispose() - vi.mocked(collectEnvironment).mockResolvedValueOnce({ ...environment, os_version_full: '10.0.26300.9458' }) + const { cpu_physical_core_count: _cores, ...partialEnvironment } = environment + vi.mocked(collectEnvironment).mockResolvedValueOnce({ ...partialEnvironment, os_version_full: '10.0.26300.9458', + missing_fields: { cpu_physical_core_count: 'timeout' }, + }) await ctx.plugin(Analytics, { enabled: true }) await ctx.productAnalytics.report({ eventName: 'api_key_save_click', timestamp: 1_800_000_000_001, attributes: {} }) await exporter.dispose() @@ -103,6 +108,8 @@ it('writes the selected event through the real exporter to an isolated collector ] }] }] }) expect(captures[0]).toContain('10.0.26300.9457') expect(captures[0]).toContain('10.0.26300.9458') + expect(captures[0]).toContain('"stringValue":"cpu_physical_core_count=timeout"') + expect(captures[0]).not.toContain('kvlistValue') expect(collectEnvironment).toHaveBeenCalledTimes(2) expect(captures[0]).not.toContain('device_id') expect(captures[0]).not.toContain('app_version') @@ -205,7 +212,7 @@ it('caches failed environment fields without discarding the event or retrying ea for (const timestamp of [1, 2]) await b.ctx.productAnalytics.report({ eventName: 'auth_page_view', timestamp, attributes: {} }) expect(collectEnvironment).toHaveBeenCalledTimes(1) expect(b.emit.mock.calls).toHaveLength(2) - expect(b.emit.mock.calls[0]![0].attributes?.missing_fields).toEqual({ os_version_full: 'timeout' }) + expect(b.emit.mock.calls[0]![0].attributes?.missing_fields).toBe('os_version_full=timeout') }) it('aborts and awaits environment sampling on disposal without emitting late results', async () => { @@ -223,3 +230,13 @@ it('aborts and awaits environment sampling on disposal without emitting late res await Promise.all([report, disposal]) expect(b.emit).not.toHaveBeenCalled() }) + + +it('encodes failed fields as a sorted scalar and omits an empty failure list', async () => { + const b = await setup(true) + vi.mocked(collectEnvironment).mockResolvedValueOnce({ ...environment, + missing_fields: { os_version_full: 'timeout', cpu_physical_core_count: 'query-failed' }, + }) + await b.ctx.productAnalytics.report({ eventName: 'auth_page_view', timestamp: 1, attributes: {} }) + expect(b.emit.mock.calls[0]![0].attributes?.missing_fields).toBe('cpu_physical_core_count=query-failed,os_version_full=timeout') +}) diff --git a/packages/client/product-analytics/tests/environment-composition.host.e2e.ts b/packages/client/product-analytics/tests/environment-composition.host.e2e.ts index 6bb08a171a..e4b1c651ae 100644 --- a/packages/client/product-analytics/tests/environment-composition.host.e2e.ts +++ b/packages/client/product-analytics/tests/environment-composition.host.e2e.ts @@ -10,8 +10,7 @@ it('adds real system metadata to the existing event through Loader and OTLP', as let captured: unknown const attribute = z.object({ key: z.string(), value: z.object({ stringValue: z.string().optional(), intValue: z.number().optional(), - kvlistValue: z.object({ values: z.array(z.unknown()) }).optional(), - }) }) + }).strict() }) const record = z.object({ eventName: z.string(), attributes: z.array(attribute) }) const batch = z.object({ resourceLogs: z.array(z.object({ scopeLogs: z.array(z.object({ logRecords: z.array(record) })) })) }) const schema = z.object({ captures: z.array(batch), reference: z.record(z.string(), z.union([z.string(), z.number()])) }) @@ -32,7 +31,7 @@ it('adds real system metadata to the existing event through Loader and OTLP', as expect(fields[key]).toEqual(typeof value === 'number' ? { intValue: value } : { stringValue: value }) } if (platform() === 'win32' || platform() === 'darwin') { - expect(fields.missing_fields).toEqual({ kvlistValue: { values: [] } }) + expect(fields).not.toHaveProperty('missing_fields') expect(fields.cpu_model?.stringValue).toBeTruthy() } if (platform() === 'win32') expect(fields.os_version_full?.stringValue).toMatch(/^\d+\.\d+\.\d+\.\d+$/) diff --git a/packages/client/product-analytics/tests/environment-query.host.spec.ts b/packages/client/product-analytics/tests/environment-query.host.spec.ts index 6b84b4ee76..a359a7c598 100644 --- a/packages/client/product-analytics/tests/environment-query.host.spec.ts +++ b/packages/client/product-analytics/tests/environment-query.host.spec.ts @@ -1,5 +1,12 @@ -import { expect, it } from 'vitest' +import { expect, it, vi } from 'vitest' +import * as childProcess from 'node:child_process' import { queryEnvironment } from '../src/environment-query.ts' +import { collectEnvironment } from '../src/environment.ts' + +vi.mock('node:child_process', async (importOriginal) => { + const actual = await importOriginal() + return { ...actual, execFile: vi.fn(actual.execFile) } +}) const options = () => ({ timeoutMillis: 5000, maxOutputBytes: 16384, signal: new AbortController().signal }) @@ -40,3 +47,24 @@ it('aborts an in-flight query and waits for child closure', async () => { lifetime.abort() expect(await result).toEqual({ kind: 'missing', reason: 'query-failed' }) }) + + +it('isolates synchronous spawn failures without rejecting or creating a child', async () => { + const spawn = vi.mocked(childProcess.execFile) + spawn.mockImplementationOnce(() => { + throw Object.assign(new Error('private spawn failure'), { code: 'ENOMEM' }) + }) + try { + expect(await queryEnvironment(process.execPath, [], options())).toEqual({ kind: 'missing', reason: 'query-failed' }) + } finally { spawn.mockClear() } +}) + + +it.runIf(process.platform === 'darwin')('matches the real macOS kernel architecture, including Intel without the arm64 OID', async () => { + const reference = await queryEnvironment('/usr/sbin/sysctl', ['-i', '-n', 'hw.optional.arm64'], options()) + expect(reference.kind).toBe('value') + if (reference.kind !== 'value') throw new Error('macOS architecture reference unavailable') + const snapshot = await collectEnvironment(options()) + expect(snapshot.os_arch).toBe(reference.stdout.trim() === '1' ? 'arm64' : 'x64') + expect(snapshot.missing_fields).not.toHaveProperty('os_arch') +}) diff --git a/packages/client/product-analytics/tests/environment.host.spec.ts b/packages/client/product-analytics/tests/environment.host.spec.ts index 6fba26822d..049595acb4 100644 --- a/packages/client/product-analytics/tests/environment.host.spec.ts +++ b/packages/client/product-analytics/tests/environment.host.spec.ts @@ -121,11 +121,11 @@ it('rejects malformed JSON without exposing raw output', async () => { expect(JSON.stringify(result)).not.toContain('private') }) -it.each(['1', '0'])('reports macOS product version and native architecture when arm64=%s', async (arm) => { +it.each(['1', undefined, '0'])('reports macOS product version and native architecture when arm64=%s', async (arm) => { vi.mocked(os.platform).mockReturnValue('darwin') vi.mocked(os.release).mockReturnValue('26.0.0') vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout: macVersion }) - .mockResolvedValueOnce({ kind: 'value', stdout: macCpu.replace('arm64: 1', `arm64: ${arm}`) }) + .mockResolvedValueOnce({ kind: 'value', stdout: arm === undefined ? 'hw.physicalcpu: 8\n' : macCpu.replace('arm64: 1', `arm64: ${arm}`) }) const result = await collectEnvironment(options()) expect(result).toMatchObject({ os_name: 'macos', os_version: 'darwin 26.0.0', os_version_full: '27.0.1', os_build: '26A434', os_arch: arm === '1' ? 'arm64' : 'x64', app_arch: process.arch, cpu_physical_core_count: 8, missing_fields: {} }) @@ -147,7 +147,7 @@ it.each(['', 'ProductVersion: garbage\nBuildVersion: private\nProductVersionExtr vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout }) .mockResolvedValueOnce({ kind: 'value', stdout: 'hw.physicalcpu: 0\nhw.optional.arm64: unknown' }) expect((await collectEnvironment(options())).missing_fields).toEqual({ - os_version_full: 'invalid-value', os_build: 'invalid-value', cpu_physical_core_count: 'invalid-value', os_arch: 'unavailable', + os_version_full: 'invalid-value', os_build: 'invalid-value', cpu_physical_core_count: 'invalid-value', os_arch: 'invalid-value', }) }) @@ -174,6 +174,21 @@ it('isolates refused synchronous OS sources', async () => { it('marks unsupported platform queries without spawning commands', async () => { vi.mocked(os.platform).mockReturnValue('linux') - expect((await collectEnvironment(options())).missing_fields).toEqual({ os_version_full: 'unsupported', os_build: 'unsupported', cpu_physical_core_count: 'unsupported' }) + expect((await collectEnvironment(options())).missing_fields).toEqual({ os_version_full: 'unsupported', os_build: 'unsupported', os_arch: 'unsupported', cpu_physical_core_count: 'unsupported' }) expect(queryEnvironment).not.toHaveBeenCalled() }) + + +it.each(['', 'ProductVersionExtra: (a)\n', 'ProductVersionExtra:\n'])('accepts an inline macOS response suffix without doubling it: %s', async (extra) => { + vi.mocked(os.platform).mockReturnValue('darwin') + vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout: macVersion.replace('27.0.1', '13.4.1 (a)') + extra }) + .mockResolvedValueOnce({ kind: 'value', stdout: macCpu }) + expect((await collectEnvironment(options())).os_version_full).toBe('13.4.1 (a)') +}) + +it('rejects conflicting macOS response suffixes', async () => { + vi.mocked(os.platform).mockReturnValue('darwin') + vi.mocked(queryEnvironment).mockResolvedValueOnce({ kind: 'value', stdout: macVersion.replace('27.0.1', '13.4.1 (a)') + 'ProductVersionExtra: (b)\n' }) + .mockResolvedValueOnce({ kind: 'value', stdout: macCpu }) + expect((await collectEnvironment(options())).missing_fields.os_version_full).toBe('invalid-value') +}) diff --git a/packages/client/product-analytics/tests/fixtures/environment.host.ts b/packages/client/product-analytics/tests/fixtures/environment.host.ts index a19cd60755..8dc08b543b 100644 --- a/packages/client/product-analytics/tests/fixtures/environment.host.ts +++ b/packages/client/product-analytics/tests/fixtures/environment.host.ts @@ -34,10 +34,14 @@ try { await analytics.report({ eventName: 'desktop_app_launch', timestamp: 1_800_000_000_000, attributes: {} }) } finally { await ctx.fiber.dispose() } if (platform() === 'darwin') { - reference.os_version_full = (await run('/usr/bin/sw_vers', ['-productVersion'])).stdout.trim() + const productVersion = (await run('/usr/bin/sw_vers', ['-productVersion'])).stdout.trim() + const { stdout: productInfo } = await run('/usr/bin/sw_vers', []) + const extra = productInfo.match(/^ProductVersionExtra:\s*(\([a-z]\))/m)?.[1] + reference.os_version_full = extra === undefined || productVersion.endsWith(extra) + ? productVersion : `${productVersion} ${extra}` reference.os_build = (await run('/usr/bin/sw_vers', ['-buildVersion'])).stdout.trim() reference.cpu_physical_core_count = Number((await run('/usr/sbin/sysctl', ['-n', 'hw.physicalcpu'])).stdout.trim()) - reference.os_arch = (await run('/usr/sbin/sysctl', ['-n', 'hw.optional.arm64'])).stdout.trim() === '1' ? 'arm64' : 'x64' + reference.os_arch = (await run('/usr/sbin/sysctl', ['-i', '-n', 'hw.optional.arm64'])).stdout.trim() === '1' ? 'arm64' : 'x64' } else if (platform() === 'win32') { const systemRoot = process.env.SystemRoot if (systemRoot === undefined) throw new Error('Windows fixture requires SystemRoot') diff --git a/tsconfig.client.json b/tsconfig.client.json index 5d7035ed6f..133b3a35f6 100644 --- a/tsconfig.client.json +++ b/tsconfig.client.json @@ -44,8 +44,8 @@ "scripts/client-bundle-css.spec.ts", "scripts/client-bundle-purity.spec.ts" ], - // A `*.host.spec.ts` covers the Host half of a split client package and - // belongs to the host aggregate, which excludes this program's `*.client.*` + // Host unit tests (`*.host.spec.ts`), e2e tests (`*.host.e2e.ts`), and fixtures + // (`*.host.ts`) belong to the host aggregate, which excludes `*.client.*` // in turn. `exclude` wins over `include`, so the test glob above stays broad. "exclude": [ "packages/client/*/tests/**/*.host.spec.ts", From 24d73cae8d23e3f0895535692f92b7b7915b68d0 Mon Sep 17 00:00:00 2001 From: winewill Date: Wed, 14 Oct 2026 16:57:28 +0800 Subject: [PATCH 026/162] feat(desktop): show complete publication approval summary --- apps/desktop/publishing/ci/README.i18n.yaml | 4 +- apps/desktop/publishing/ci/README.md | 2 +- apps/desktop/publishing/ci/README.zh.md | 2 +- .../scripts/desktop-publication-summary.ts | 70 +++++++++++++++++++ apps/desktop/scripts/publish-desktop-ci.ts | 4 +- .../desktop-publication-preflight.spec.ts | 8 +++ .../tests/desktop-publication-summary.spec.ts | 46 ++++++++++++ .../tests/expected/publication-mac-arm64.md | 47 +++++++++++++ .../tests/expected/publication-production.md | 41 +++++++++++ .../tests/expected/publication-win-x64.md | 46 ++++++++++++ 10 files changed, 264 insertions(+), 6 deletions(-) create mode 100644 apps/desktop/scripts/desktop-publication-summary.ts create mode 100644 apps/desktop/tests/desktop-publication-summary.spec.ts create mode 100644 apps/desktop/tests/expected/publication-mac-arm64.md create mode 100644 apps/desktop/tests/expected/publication-production.md create mode 100644 apps/desktop/tests/expected/publication-win-x64.md diff --git a/apps/desktop/publishing/ci/README.i18n.yaml b/apps/desktop/publishing/ci/README.i18n.yaml index d89cc5e4f8..72053a7e6f 100644 --- a/apps/desktop/publishing/ci/README.i18n.yaml +++ b/apps/desktop/publishing/ci/README.i18n.yaml @@ -18,8 +18,8 @@ en: 990ff7e56dd6005a zh: 18f9dc23c520dc8c /desktop-ci-publication/package-and-publish: - en: c17307a96efdde3a - zh: 53ba328b149ed407 + en: 1aaab994b182e893 + zh: 54bc0689b1f3a7c5 /desktop-ci-publication/recovery-and-records: en: e92bfad06162443e zh: 6c3e3d8ece673217 diff --git a/apps/desktop/publishing/ci/README.md b/apps/desktop/publishing/ci/README.md index f3d50a61a3..09920d83c1 100644 --- a/apps/desktop/publishing/ci/README.md +++ b/apps/desktop/publishing/ci/README.md @@ -39,7 +39,7 @@ Select the platform's `Package Desktop` workflow, preset, source, channel and ve Packaging only retains files. After success, open **Publish Desktop · COS → Run workflow** on `master`, enter `package_run_id` from the packaging run URL, and optionally enter `package_run_attempt`. Blank selects the latest attempt, which must have succeeded; it never falls back to an older success. An explicit attempt selects that exact successful build. Windows must be signed; macOS must finish notarization and stapling. GitHub job evidence selects the original runner; its local manifest supplies the preset, version, source and channel before approval. No installer or request artifact is downloaded. Retained packages with the old `publishRequested=false` flag remain eligible; missing or changed local files stop publication. -Before uploading, the publisher hashes all retained files, checks version and destination metadata, and rechecks the platform signatures. macOS verifies the DMG and extracts the ZIP to verify the app's ticket; Windows checks Authenticode, the configured signer thumbprint and the timestamp. Preparation writes an approval summary with version, channel, platform, deployment, source ref/SHA, package run/attempt, object keys and proposed tag. Both publishing jobs wait for **Review deployments → Approve and deploy**, then recheck the approved manifest and upload plan digest. Changed bytes or destinations reject. Approval cannot edit a version; reject and rebuild instead. Both deployments upload missing binaries, verify remote bytes, then update YAML. There is no files-only mode in this workflow. Production finally creates `desktop-v` at the recorded packaging commit; an existing tag must name the same commit. +Before uploading, the publisher hashes all retained files, checks version and destination metadata, and rechecks the platform signatures. macOS verifies the DMG and extracts the ZIP to verify the app's ticket; Windows checks Authenticode, the configured signer thumbprint and the timestamp. Preparation writes a Chinese approval summary with separate version, channel, platform and deployment fields, the retained test release ID, full installer/feed URLs, a link to the original packaging attempt, source ref/SHA, platform verification results and proposed tag. Folded details include the COS bucket and file sizes/SHA-256 of the upload bytes, including rewritten YAML. Public-repository summaries expose the test prefix and download URLs; the prefix is not authentication. Both publishing jobs wait for **Review deployments → Approve and deploy**, then recheck the approved manifest and upload plan digest. Changed bytes or destinations reject. Approval cannot edit a version; reject and rebuild instead. Both deployments upload missing binaries, verify remote bytes, then update YAML. There is no files-only mode in this workflow. Production finally creates `desktop-v` at the recorded packaging commit; an existing tag must name the same commit. diff --git a/apps/desktop/publishing/ci/README.zh.md b/apps/desktop/publishing/ci/README.zh.md index 97bb986b67..2362c427b7 100644 --- a/apps/desktop/publishing/ci/README.zh.md +++ b/apps/desktop/publishing/ci/README.zh.md @@ -39,7 +39,7 @@ Environment 审批限制此工作流，不限制同一机器账户运行的任 打包只保留文件。成功后，在 `master` 上打开 **Publish Desktop · COS → Run workflow**，填写打包运行 URL 中的 `package_run_id`，并可填写 `package_run_attempt`。留空选择最新尝试，且该尝试必须成功，不会回退到较早的成功尝试；显式填写则选择对应的成功构建。Windows 必须完成签名，macOS 必须完成公证和 stapling。GitHub job 证据确定原 runner，本机清单在审批前提供 preset、版本、源码和通道。不会下载安装包或请求 artifact。旧清单的 `publishRequested=false` 不影响发布资格；本机文件缺失或变化会停止发布。 -上传前，发布器校验全部留存文件的哈希、版本及目标信息，再复核平台签名。macOS 校验 DMG，并解压 ZIP 校验 app 票据；Windows 校验 Authenticode、配置的签名证书指纹及时间戳。准备阶段生成审批摘要，列出版本、通道、平台、环境、源码 ref/SHA、package run/attempt、对象路径及拟创建的 tag。两种发布 job 均等待 **Review deployments → Approve and deploy**，随后重新核对已批准的 manifest 与上传计划摘要。文件或目标变化会被拒绝。审批不能修改版本；应拒绝并重新打包。两种环境都上传缺失文件、校验远端字节，再更新 YAML。此工作流没有只上传文件的模式。production 最后为记录的打包提交创建 `desktop-v`；已有 tag 必须指向同一提交。 +上传前，发布器校验全部留存文件的哈希、版本及目标信息，再复核平台签名。macOS 校验 DMG，并解压 ZIP 校验 app 票据；Windows 校验 Authenticode、配置的签名证书指纹及时间戳。准备阶段生成中文审批摘要，分别展示版本、通道、平台和环境，以及留存的 test release ID、安装包与升级 YAML 的完整 URL、原打包尝试链接、源码 ref/SHA、平台校验结果及拟创建的 tag。折叠详情包含 COS 存储桶，以及实际上传字节的大小和 SHA-256（包括改写后的 YAML）。公开仓库的摘要会公开测试前缀和下载 URL；前缀不提供访问鉴权。两种发布 job 均等待 **Review deployments → Approve and deploy**，随后重新核对已批准的 manifest 与上传计划摘要。文件或目标变化会被拒绝。审批不能修改版本；应拒绝并重新打包。两种环境都上传缺失文件、校验远端字节，再更新 YAML。此工作流没有只上传文件的模式。production 最后为记录的打包提交创建 `desktop-v`；已有 tag 必须指向同一提交。 diff --git a/apps/desktop/scripts/desktop-publication-summary.ts b/apps/desktop/scripts/desktop-publication-summary.ts new file mode 100644 index 0000000000..19b55a8f7d --- /dev/null +++ b/apps/desktop/scripts/desktop-publication-summary.ts @@ -0,0 +1,70 @@ +/** Render the verified publication plan for GitHub's native deployment reviewers. */ +import { createHash } from 'node:crypto' +import { createReadStream } from 'node:fs' +import type { DesktopRetainedSource } from './desktop-release-store.ts' +import type { DesktopUploadPlan } from './desktop-upload-plan.ts' + +function text(value: string): string { + return value.replace(/[&<>"'|`\r\n]/gu, character => `&#${character.charCodeAt(0)};`) +} + +function code(value: string): string { return `${text(value)}` } +function row(label: string, value: string): string { return `| ${label} | ${value} |` } +function link(url: string, label = url): string { return `${text(label)}` } + +/** + * Render Chinese approval details after package, destination and signature verification succeeds. + * @param source Verified retained packaging identity. + * @param plan Validated upload plan; YAML contents are the bytes that will be uploaded. + * @param origin Validated HTTPS download origin from the publishing configuration. + * @param approval Digest bound to the approved manifest and upload plan. + * @returns GitHub-flavored Markdown with final URLs and upload-byte fingerprints, without credentials or local paths. + */ +export async function desktopPublicationSummary( + source: DesktopRetainedSource, plan: DesktopUploadPlan, origin: string, approval: string, +): Promise { + const url = (key: string) => `${origin}/${key.split('/').map(encodeURIComponent).join('/')}` + const files = [] + for (const file of plan.artifacts) { + const hash = createHash('sha256') + let size = 0 + const bytes = file.contents === undefined ? createReadStream(file.path) : [Buffer.from(file.contents)] + for await (const chunk of bytes) { size += chunk.length; hash.update(chunk) } + files.push(row(link(url(file.key), file.filename), `${size} 字节 | ${code(hash.digest('hex'))}`)) + } + const installerExtension = plan.target === 'win-x64' ? '.exe' : '.dmg' + const installers = plan.artifacts.filter(file => !file.channelMetadata && file.filename.endsWith(installerExtension)) + const feeds = plan.artifacts.filter(file => file.channelMetadata) + const runUrl = `https://github.com/${source.repository}/actions/runs/${source.runId}/attempts/${source.attempt}` + const environment = plan.environment === 'test' ? 'desktop-test' : 'desktop-production' + const tag = plan.environment === 'production' ? code(`desktop-v${plan.version}`) : '不创建（test 环境）' + return [ + '', '## 发布前确认', '', + `核对以下信息后，点击 **Review deployments**，选择 ${code(environment)} 并点击 **Approve and deploy**；信息有误请点击 **Reject**。`, '', + '| 项目 | 发布信息 |', '| --- | --- |', + row('发布环境', code(plan.environment)), + row('版本号', code(plan.version)), + row('升级通道', code(source.channel)), + row('平台', code(plan.target)), + ...(plan.environment === 'test' ? [row('测试 release ID', `${code(source.testReleaseId!)}
测试下载路径的随机前缀，用于降低地址被猜到的可能性；连续升级测试需使用相同值。`)] : []), + row('下载地址预览', `${installers.map(file => link(url(file.key))).join('
')}
发布后的目标地址，审批前不保证可下载。`), + row('升级 YAML', `${feeds.map(file => link(url(file.key))).join('
')}
本次发布将更新此文件；审批前可能不存在或仍指向旧版本。`), + row('打包来源', `${link(runUrl, `查看打包运行 #${source.runId} · 第 ${source.attempt} 次尝试`)}
源码：${code(`${source.sourceRef} @ ${source.commit}`)}
打包机器：${code(source.runnerName)}`), + row('签名与公证', plan.target === 'win-x64' ? 'EXE 签名校验通过' : 'DMG 校验通过
ZIP 内应用校验通过'), + row('Git tag', tag), '', '### 批准后将执行', '', + '1. 重新校验产物与发布目标；若与审批摘要不一致则停止。', + '2. 上传缺少的文件，跳过内容一致的已有文件；发现内容冲突则停止。', + '3. 更新上述升级 YAML，并保存发布记录。', + ...(plan.environment === 'production' ? ['4. 为实际打包提交创建上述 Git tag；若 tag 创建失败，记录“已发布，待补 tag”。'] : []), '', + ...(plan.environment === 'test' ? ['**本次只发布 test，不修改 production。**', ''] : []), + '
', '技术详情与文件清单', '', + `COS 存储桶：${code(plan.bucket)}`, '', + ...(plan.environment === 'test' ? [ + 'release ID 取自原打包产物清单。打包时优先使用表单 `test_release_id`；留空则使用机器 `presets/test.env` 中的 `DOWNLOAD_TEST_RELEASE_ID`。发布时不重新选择。', '', + ] : []), + `审批摘要 SHA-256：${code(approval)}`, '', + `共 ${plan.artifacts.length} 个文件；大小及 SHA-256 对应实际上传字节，包括更新后的 YAML 内容。`, '', + '| 文件 | 大小 | SHA-256 |', '| --- | --- | --- |', ...files, '', '
', '', + '此摘要在公开仓库中可公开查看，包含下载路径及 test 发布的 release ID。随机前缀不能替代访问鉴权。', '', + ].join('\n') +} diff --git a/apps/desktop/scripts/publish-desktop-ci.ts b/apps/desktop/scripts/publish-desktop-ci.ts index 696ceea2cd..eb4018858f 100644 --- a/apps/desktop/scripts/publish-desktop-ci.ts +++ b/apps/desktop/scripts/publish-desktop-ci.ts @@ -12,6 +12,7 @@ import { desktopPublishCos } from './desktop-publish-cos.ts' import { createDesktopPublicationRecord, wasDesktopAttemptPublished, withDesktopChannelLock } from './desktop-publish-record.ts' import { publishDesktopRelease } from './desktop-publish.ts' import { desktopPublicationFailure } from './desktop-publication-failure.ts' +import { desktopPublicationSummary } from './desktop-publication-summary.ts' import { publishDesktopTag } from './desktop-publish-tag.ts' const routeSchema = z.object({ runId: z.string(), attempt: z.string(), commit: z.string(), runnerName: z.string(), @@ -62,8 +63,7 @@ if (process.argv[1] !== undefined && resolve(process.argv[1]) === import.meta.fi if (mode === '--check') { if (env.GITHUB_OUTPUT) await appendFile(env.GITHUB_OUTPUT, `approval=${approval}\nlabel=${plan.version} → ${manifest.source.channel} / ${plan.target} / ${plan.environment}\n`) if (env.GITHUB_STEP_SUMMARY) await appendFile(env.GITHUB_STEP_SUMMARY, - `\n## Confirm Desktop publication\n\nVersion: **${plan.version}**; channel: **${manifest.source.channel}**; environment: **${plan.environment}**; platform: **${plan.target}**\n\nSource: ${manifest.source.sourceRef} @ ${route.commit}\n\nPackage: ${route.runId}, attempt ${route.attempt}\n\nTag: ${plan.environment === 'production' ? `desktop-v${plan.version}` : 'none (test)'}\n\nApproved package SHA-256: ${approval}\n\n` + - plan.artifacts.map(file => `- ${file.channelMetadata ? 'Feed' : 'Artifact'}: ${file.key}`).join('\n') + '\n') + await desktopPublicationSummary(manifest.source, plan, config.origin, approval)) if (env.GITHUB_OUTPUT && config.githubApp) await appendFile(env.GITHUB_OUTPUT, `app-client-id=${config.githubApp.clientId}\napp-key-file=${config.githubApp.privateKeyFile}\n`) process.stdout.write(`desktop publish: verified ${plan.target} ${plan.version}; no credentials read or COS requests made\n`) diff --git a/apps/desktop/tests/desktop-publication-preflight.spec.ts b/apps/desktop/tests/desktop-publication-preflight.spec.ts index f4aaedfb8c..29c07ea347 100644 --- a/apps/desktop/tests/desktop-publication-preflight.spec.ts +++ b/apps/desktop/tests/desktop-publication-preflight.spec.ts @@ -4,6 +4,7 @@ import { tmpdir } from 'node:os' import { join } from 'node:path' import { expect, it } from 'vitest' import { retainDesktopCi } from '../scripts/retain-desktop-ci.ts' +import { desktopPublicationSummary } from '../scripts/desktop-publication-summary.ts' import { prepareDesktopPublication } from '../scripts/publish-desktop-ci.ts' it.each(['mac-arm64', 'win-x64'] as const)('preflights a complete retained %s build without credentials or source checkout manifests', async (target) => { @@ -51,6 +52,13 @@ it.each(['mac-arm64', 'win-x64'] as const)('preflights a complete retained %s bu expect(prepared.plan.version).toBe('1.0.0') expect(prepared.plan.artifacts.every(file => file.path.startsWith(retained))).toBe(true) expect(signatureChecks).toBe(1) + const summary = await desktopPublicationSummary(prepared.manifest.source, prepared.plan, prepared.config.origin, 'c'.repeat(64)) + await expect(summary).toMatchFileSnapshot(`./expected/publication-${target}.md`) + const feed = prepared.plan.artifacts.find(file => file.channelMetadata)! + expect(summary).toContain(createHash('sha256').update(feed.contents!).digest('hex')) + expect(summary).toContain('https://github.com/owner/repo/actions/runs/10/attempts/2') + expect(summary).not.toContain(root) + expect(summary).not.toContain('absent-credentials') await expect(prepareDesktopPublication({ ...env, DESKTOP_APPROVED_PACKAGE: 'wrong' }, signature)).rejects.toThrow(/changed after approval/u) await expect(prepareDesktopPublication({ ...env, RUNNER_NAME: 'other-machine' }, signature)).rejects.toThrow(/different machine/u) await writeFile(join(retained, `${base}.${extension}`), 'changed') diff --git a/apps/desktop/tests/desktop-publication-summary.spec.ts b/apps/desktop/tests/desktop-publication-summary.spec.ts new file mode 100644 index 0000000000..4f467f5ca7 --- /dev/null +++ b/apps/desktop/tests/desktop-publication-summary.spec.ts @@ -0,0 +1,46 @@ +import { createHash } from 'node:crypto' +import { expect, it } from 'vitest' +import { desktopPublicationSummary } from '../scripts/desktop-publication-summary.ts' +import type { DesktopRetainedSource } from '../scripts/desktop-release-store.ts' +import type { DesktopUploadPlan } from '../scripts/desktop-upload-plan.ts' + +const source: DesktopRetainedSource = { + repository: 'owner/repo', runId: '20', attempt: '3', commit: 'a'.repeat(40), workflowCommit: 'b'.repeat(40), + sourceRef: 'dsh-v1.0.0', channel: 'latest', runnerName: 'windows-publisher', target: 'win-x64', + preset: 'production', productVersion: '1.0.0', version: '1.0.0', testReleaseId: null, qualification: 'signed', +} +const plan: DesktopUploadPlan = { + environment: 'production', target: 'win-x64', version: '1.0.0', bucket: 'production-1234567890', + publicUrl: 'https://download.example.com/dsh-desk/feeds/win-x64/', secretIdEnvName: 'SECRET_ID', secretKeyEnvName: 'SECRET_KEY', + artifacts: [ + { path: '/not-read/installer.exe', filename: 'Harness 1.0.0.exe', key: 'dsh-desk/bin/win-x64/latest/Harness 1.0.0.exe', + contents: 'installer bytes', contentType: 'application/octet-stream', channelMetadata: false }, + { path: '/not-read/latest.yml', filename: 'latest.yml', key: 'dsh-desk/feeds/win-x64/latest.yml', + contents: 'version: 1.0.0\n', contentType: 'application/yaml', channelMetadata: true }, + ], +} + +it('shows production destinations and tag after approval with upload-byte fingerprints', async () => { + const summary = await desktopPublicationSummary(source, plan, 'https://download.example.com', 'c'.repeat(64)) + await expect(summary).toMatchFileSnapshot('./expected/publication-production.md') + expect(summary).toContain('/latest/Harness%201.0.0.exe') + expect(summary).toContain('https://download.example.com/dsh-desk/feeds/win-x64/latest.yml') + expect(summary).toContain(createHash('sha256').update('version: 1.0.0\n').digest('hex')) + expect(summary).not.toContain('测试 release ID |') + expect(summary).not.toContain('SECRET_KEY') + expect(summary).not.toContain('/not-read/') +}) + +it('escapes runner display text without adding Markdown rows or HTML elements', async () => { + const summary = await desktopPublicationSummary({ ...source, runnerName: 'runner|\n