Turtle
d1e22a7e24
* feat(preset): declare Agent compositions in profile YAML * fix(preset): share scope binding cleanup and refresh module graph * fix(preset): use host syntax colors in the YAML editor * fix(preset): write new profile declarations as block YAML * fix(preset): align eager activation with browser and worker surfaces * fix(ui-agent-preset): retain the existing preset card layout * fix(preset): keep rows waiting for a Host service mounted and re-audit after settlement A preset row whose Host provider was still activating at declaration time was reported as waiting for a missing service, and the registry recorded the definition as broken for the rest of the process. Split the mount audit into failed and pending rows: failures and root-realm leaks still reject the mount, while pending rows stay mounted and every read and binding re-audits them after the Host Loader tree settles. Remove the accidental self-dependency of @deepseek-ai/dsh-web-app and its lockfile self-link. Archive the copy-only preset authoring note, which the declarative-preset note fully supersedes, and update the ralph note's restore recipe to the profile-patch mechanism. * chore(lint): re-key and prune the unknown-cast baseline after the master merge * refactor(web-app): ship each Agent preset as its own bundle patch file `dsh.bundle.patch` accepts an ordered file list, so the four shipped preset declarations move out of the 1000-line web patch into `presets/<id>.patch.yml`, one insert per file, applied after the host patch. The host patch keeps the preset registry row. Gates and test scaffolds that read the web layer resolve every patch file through `bundlePatchPaths` instead of one hardcoded path. * refactor(preset): drop the save notice injected into live Agents The notice duplicated the save result the Web editor already shows, and after the producer-owned message sources it would have needed its own persisted source kind. Saving now only persists and reports the activation diagnostic; the persistence catalog is unchanged from master. * docs: regenerate the module graph after dropping the registry's llm dependency * test(ui-agent-preset): cover the guide dialog's keyboard handling on the declarative section * fix(preset): address the remaining review comments Apply the lowercase id rule only when creating a declaration, so presets declared with other ids stay editable. Export `isBuiltInPreset` from the display fold and key the client card groups on it. Remove the dead `error` and `retry` locale keys, the JSDoc orphaned by the `trust` removal, the `trust` fields in the plugin-inventory test fixtures, and the stale `dsh-agent-presets` package name in comments and READMEs. * refactor(preset): make the Web preset page view-and-select only Remove the child-plugin YAML editor, the `agent_preset` tool and the registry's read/save Remote methods. Presets are declared and changed only by bundle patches; the settings page keeps the roster, the new-task default and the mode help, and its question icon explains that and offers to start a Creator-mode task. The `editing-cordis-compositions` skill and the Creator persona describe the bundle-based path, and the authored skill-load Session and Web snapshots are refreshed for it. * docs(skill): make the composition skill self-sufficient for preset authoring State where the shipped declarations live, list every declaration field, inline a complete minimal bundle and an override example, and name the verification calls, so Creator mode no longer reads DSH package sources to infer the format. Refresh the authored skill-load Session for the new text. * test(tools): drop the removed agent_preset tool from the catalog expectation
deepseek-harness-runtime-bin
English | 中文
Platform runtime wheel for the DeepSeek Harness Python SDK. It packages the normal dsh CLI and its closed Node dependency tree into a native executable, so SDK use requires no system Node.js. This package publishes wheels only.
Installed commands and artifacts
The wheel installs a dsh console command and the deepseek_harness_runtime Python module. dsh forwards its arguments to the bundled executable and requires a non-empty DSH_HOME; it never falls back to ~/.dsh.
Production executables are named deepseek-harness-sdk-runtime-<platform>-<arch> under the module's runtime/ directory; Windows uses the .exe suffix. Linux and macOS wheels include a target-native -rg sidecar, Windows includes -rg.exe, and macOS also includes -spawn-helper for node-pty. Published targets are Linux x64, Linux arm64, macOS arm64, macOS x64, and Windows x64. The wheel tag and payload must match exactly; no Windows arm64 wheel is published.
Each target also requires <executable-stem>-office/, where the stem excludes .exe. This directory contains the complete installed Office packages and their dependencies, preserving engine resources, manifests, licenses, source inventories, and helper permissions. Copy this directory together with the executable. A missing target engine fails the sidecar build with its npm package name and target platform/architecture.
Each wheel also includes <platform>-<arch>/primary-runtime/ (CPython and the locked Office Python libraries) and sibling office-skills/ (three default workflows and their shared checker). These are ordinary relocatable files, not bytes embedded in the executable. The shared builder selects target-native archives for all five wheel targets and executes its smoke on the native build host. Packaging and installed-runtime lookup reject missing resources, wrong-platform metadata, and lost Python executable permissions. The short platform directory avoids repeating the executable name in Python DLL paths on Windows.
The packaged bootstrap supplies DSH_BUNDLED_PRIMARY_RUNTIME as a carrier default. The sdk profile uses it when DSH_PRIMARY_RUNTIME is unset; an explicit path overrides it, and an empty string disables the query and Office provider. External payloads retain the primary-runtime/ plus sibling office-skills/ layout. The SDK reads Python in place without copying it into DSH_HOME. Source and dev-only Node carriers have no bundled default; they use an explicit DSH_PRIMARY_RUNTIME.
Skill selection is independent of delivery: project, custom-directory, and user filesystem skills override same-name bundled skills. An SDK patch can disable only the Office provider while retaining the Python query:
- id: skill-office
disabled: true
To replace the three Office workflows and shared checker as a set, patch skill-office.config.assetRoot to another absolute resource directory. Use the filesystem skill provider for arbitrary skill collections. Configuration patches apply at process startup; changing skills does not require rebuilding the runtime wheel or Python environment.
Repository builds also materialize a dev-only runtime/node/ carrier. It runs node runtime/node/node_modules/@deepseek-ai/dsh/lib/bin.js on system Node 22.19 or newer. It is never selected automatically and is excluded from wheels and sdists.
Both carriers execute the same dsh grammar and shipped profiles, including the standalone sdk-minimal tree and the full web profile with its frontend assets. The private dsh-python-runtime-closure manifest defines the packaged dependency closure; there is no Python-specific Node application or checked-in default cordis.yml.
Python module API
bundled_package_dir() -> Pathreturns the installed module-data root and verifies its release metadata.bundled_runtime_path() -> Pathreturns the current platform executable and verifies required sidecars.resolve_bundled_launch_args(mode=None) -> tuple[str, ...]returns the executable argv by default. Explicitmode="node"orDSH_RUNTIME_MODE=nodeselects the repo-only Node carrier.main()implements the installeddshconsole command and rejects an absent or blankDSH_HOME. On Windows it waits for the bundled process with inherited standard streams and forwards its exit status; on POSIX it replaces the Python process.
Unsupported platforms and missing executables or sidecars raise FileNotFoundError with the build and installation routes. Unknown runtime modes raise ValueError.
Packaged profile resolution
dsh initializes shipped profiles under the explicit home, composes their bundle patches, and loads bundled plugins from the executable's virtual filesystem. Runtime resolution uses an in-memory generation instead of disk symlinks or proxy packages. Fallback imports use recorded declaring-package paths, including paths inside the executable's virtual filesystem, so built-in rows and external plugin peers share the bundled Cordis/module instance. Native shared libraries and Windows ConPTY addons are packaged with native addons, while ripgrep and the macOS PTY helper remain executable sidecars.
The Python bootstrap resolves the Office kit from its adjacent directory so native helpers and URL Workers use real filesystem paths. The kit owns engine selection and validation; the Python bootstrap adds no runtime download or compilation.
External profile management uses dsh plugin --profile <name> .... That command requires pnpm on PATH; ordinary SDK/profile execution does not.
Build and distribution
Production deployment permits unused workspace patches for packages outside the runtime closure; patches for included packages must still apply successfully. This exception is confined to the deploy command; repository installation still rejects unused patches.
From the repository root, pnpm exec tsx scripts/build-exe-for-python-sdk.ts verifies the closure, builds packages, deploys a symlink-free tree, packages the selected target, and syncs the executable and sidecars into this module. scripts/build-python-release.py stages release-shaped wheels at the root repository version and pins deepseek-harness-sdk to the exact runtime version.
The installed-wheel smoke creates a clean virtual environment outside the checkout, proves the installed distribution and executable identities, then exercises default and customized SDK profiles, external plugins, MCP, native tools, direct JSON-RPC, committed snapshots, and the real provider on trusted runs. Its Office scenario relocates the complete target payload and converts DOCX with the required platform engine: the target’s declared native engine, or WASM when no native engine is declared. See the Python contributor workflow and installed-wheel testing decision.