Turtle
38c7a9fbf4
* feat(release): publish package source, declaration maps, and documentation Every release member publishes its TypeScript source and the declaration maps that resolve into it, and @deepseek-ai/dsh publishes the repository Markdown documentation staged from docs/ at prepack time. The ./src/* export most members already declared now resolves for an installed consumer, and a consumer editor follows a declaration to the implementation. Bundled JavaScript maps stay unpublished; the vendored family and the Desktop runtime file policy are unchanged. Rationale: .agents/notes/implemented/process/2026-10-08-published-source-and-docs.md * feat(release): publish JavaScript maps and compress bundles Every emitted JavaScript map ships beside the bundle it describes, and the workspace build and the client preset compress bundles without mangling identifiers. The documentation payload now mirrors the website publication manifest and is removed again by `postpack`; Desktop and the Python executable exclude the published source, documentation, and maps; `clean` removes a staged copy. Typert's publication contract gains the Remote declaration map. * fix(release): keep node bundles readable for the worker image The browser worker image lowers `yield*` only as a whole statement and rewrites `for await` loops by their source layout, so compressed node-side entries either fail that lowering or lower to unparseable code. Only the browser bundles keep compression, which removes about a third of their text. The bundle-purity assertion now matches the module-loader call regardless of the quote style compression picks, and the release-pack fixture exercises the installed package's new documentation requirement. * feat(release): publish the whole documentation tree `@deepseek-ai/dsh` copies `docs/` in full instead of the website's page set: the maintainer documents, the generated migration records, the translation guidelines, and the `*.i18n.yaml` consistency records ship with the installed package, so the documentation's own links resolve without a checkout. * feat(release): publish the upgrade guides with the documentation The staged documentation keeps the website's page set and adds every `docs/upgrade-guide/` document, which is what an installed reader consults when a released change breaks something. The rest of the maintainer tree stays out. * test(release): expect source and maps in master's new payload cases * feat(release): keep installed docs and source readable in Desktop Desktop keeps the published src/ and docs/ trees inside app.asar, and every model shell call receives DSH_INSTALL_DIR and DSH_NODE_EXECUTABLE so an agent can read them through Electron's Node mode. Staging also generates docs/upgrade-guide/README.md, an index of the upgrade guides grouped by the release they upgrade from. Minified client bundles print string literals with backticks: the worker image packer now recognizes those chunk requests, client license banners are prepended after minification so their text survives, and the affected artifact tests accept either quote. * test(desktop): read archived docs through the Windows PowerShell form Regenerate the config catalog for the shifted shell-env source line. * test(desktop): wait for the GUI-subsystem Electron process from PowerShell * build(release): publish source and maps from the message-quotes package * build(release): publish source and maps from packages new on master * build(release): publish source and maps from the tool-office package
description, kind
| description | kind |
|---|---|
| The Web feedback surface: the Like/Dislike pair in the finalized assistant message's action row, the feedback dialog behind both ratings and `/feedback`, and its acknowledgement and failure toasts; for users and maintainers of the feedback experience. | package-reference |
@deepseek-ai/dsh-client-ui-message-feedback
English | 中文
Summary
This package is the Web GUI's feedback surface: the Like/Dislike pair in the finalized assistant message's action strip, the feedback dialog with its acknowledgement and failure toasts in the composer overlay, and a decoration that opens the dialog from a bare /feedback. Like and Dislike both open the dialog, which collects a category and an optional description before recording the selected rating. One surface per Session backs every entry, so a single list read seeds the whole transcript and one dialog serves the Session and its messages. Ratings, categories, and notes are log-only Session events that never enter model context.
Table of Contents
- Use this package
- Understand the implementation
- Further Exploration
- Model Experience
- Known Limitations and Deferred Work
- Dev Note
Use this package
Mount this plugin alongside ui-conversation and ui-commands; the Like/Dislike pair then appears in the action row of each turn's closing assistant message, between copy and branch, and Feedback in the composer menu opens the same dialog. Mounting session-log-export also adds Feedback to the Session Header's more-actions menu. A recorded rating shows the filled glyph and stays visible without hover. Like and Dislike both open the dialog: seven category chips and a detail box are optional, and Submit records the selected judgment with whatever was filled in before the toast thanks the user; the conversation log travels with every feedback event. Clicking the recorded rating retracts it without opening the dialog. A bare /feedback, picked from the menu or typed and sent without text, opens the same dialog for the Session; /feedback <text> keeps the Host command path and its acknowledgement row.
Failures
A rating or list-load failure shows inline in the row; a submission failure shows in a warning toast while the dialog stays open so the draft can be corrected. Only finalized messages reach the message entry — an interruption-frozen partial carries no messageId and therefore no feedback controls.
Access: the rating buttons render only while the current Peer may call messageFeedback/put on the Session; the /feedback decoration has no Client check, and the Host refuses a submission without access.
Understand the implementation
Implementation internals — click to expand
The package contributes the feedback entry (order 10) of conversation.chat.assistant-actions, declared by ui-conversation and rendered inside the finalized assistant message's IconActions row, and the feedback-dialog entry (order 2) of conversation.input.overlay, which renders the Modal and Toast primitives through body portals and centers the toast over the composer card it mounts inside. The feedbackUi.openSession(sessionId) Client service opens that same Session-scoped dialog without recording feedback; the Header menu and command decoration both use it. The /feedback decoration is an action registered through ctx.commandUi.decorate, so a menu pick or a bare Enter consumes the trigger token and opens the dialog while an argued line still reaches the Host command.
Per Session, one MessageFeedbackController backs every message control and one FeedbackDialogController owns the dialog draft, the submission, and the toast sequence. The message controller reads messageFeedback.list once, deferred to the first hover or focus rather than fired on mount, and serializes mutations so each carries the version last observed; a version-conflict reply carries the authoritative item and reconciles the view without refetching. Before either rating action proceeds, the row checks the committed item: the matching rating calls retract, which rechecks the rating in the serialized queue and becomes a no-op after a concurrent change, while any other state opens the dialog with the requested rating. The dialog controller submits by target: a message target puts that rating with the dialog's note and category through the message controller, and the Session target records through ctx.remote.sessionFeedback. Success closes the draft and raises the acknowledgement toast; a late success from a superseded draft raises that toast without closing the new draft; a failure keeps the draft open and raises a longer-lived warning toast.
Further Exploration
Read these pages when the feedback surface is not enough. They move from the browser strip to the Session-log backends and the conversation shell.
- dsh-message-feedback — the Session-log backend that owns per-item compare-and-set and persistence.
- dsh-command-feedback — the
/feedbackcommand, thesessionFeedbackRemote, and the category taxonomy. - ui-commands — the command decoration contract the
/feedbackrow goes through. - ui-conversation — declares the assistant-actions strip and the composer overlay.
- Client package map — adjacent browser UI packages.
Model Experience
None, as ratings, categories, and notes are log-only events, not model input. Optional Session-log delivery uses request metadata rather than model context.
KV Cache effect
None; feedback mutations leave the model-visible history unchanged.
Known Limitations and Deferred Work
These limits define the current feedback surface. They are current package constraints, not a general rating comparison or a task backlog.
- Note size is a Host policy — the deployment configures
maxNoteBytes(8192 in the Web bundle) and the Host rejects an oversized note withnote-too-large. The dialog does not pre-check the limit, so an oversized description for a message fails on submit rather than while typing; a Session remark has no bound. - No cross-tab push — a second tab's rating becomes visible on reconnect or on the next conflict reply, not immediately; the controller does not consume feedback log events.
- Chat view only — the trajectory and waterfall views render no feedback controls even though their assistant nodes carry the same
messageId.
Dev Note
Working context for maintainers — click to expand
None.