Why: U4 read-only experience - surface.updated events project into a
surfaces slice with a strict revision guard (stale/duplicate events
never overwrite; server snapshot is the single owner), preview tabs
gain a surface: kind that renders SurfaceView (doc-refs bridge into
artifact previews), a new page auto-opens a tab without stealing the
user's active non-surface tab, and opening a conversation reloads the
authoritative surface list so refresh/断线 restores pages.
Evidence: tsc clean; vitest 87/87 including new surface projection
tests (stale/duplicate/multi-surface); UI SurfaceView suite green in
its repo.
Impact: Frontend-only; no backend or contract changes.
Refs: Core PRs #27/#28; UI PRs #42/#43/#44; Dev task
feature/generative-ui-u4
Why: Core PR #26 added the node id/source fields to ts/types.ts (the
single authority); the mirror is regenerated so the sync header and
content match make contracts output exactly (manual edits drift).
Evidence: File generated by the make contracts copy steps; combined
verification in the Dev workspace after re-pinning.
Impact: Header banner plus the already-reviewed id/source fields.
Refs: Core PR #26; Dev task feature/generative-ui-u2
Why: Core PR #25 adds optional id/source to UINode for the Generative
UI Surface layer; the TypeScript contract authority mirrors core types.
Evidence: Type-only additive change; combined verification runs in the
Dev workspace make test.
Impact: No component or behavior change.
Refs: Agent-System-Core #25; Dev task feature/generative-ui-u2
Why: Removing the artifacts builtin shifts the shortcut ranking, so the
workspaces tests now expect agents at cmd-1; Workbench no longer selects
previewOpen in its outer scope (FilesTab owns that read).
Evidence: npx vitest run 85/85 passed; npm run build (tsc -b && vite
build) passed in the task worktree (SELF).
Impact: test/workspaces.test.ts expectations; Workbench.tsx selector
cleanup. No behavior change beyond the committed layout rework.
Why: The three-column workbench redesign converges entries and hierarchy:
project switching tops the left rail (identity shown once), sessions and
search stay centered, files/workspaces/account sink to the bottom. The
right column becomes a Work/Files third column that absorbs the preview
pane, which is no longer an independent fourth region; the artifacts
workspace overlay folds into the Files entry, and stop control
consolidates into the input area with a light fast/slow-lane status hint.
Evidence: make test-frontend NOT_RUN at commit time; the run and its
outcome are recorded in the Dev task card before integration.
Impact: App.tsx layout and divider width tiers (work 280-560, files with
open preview 420-1100); store UI state rightTab work/files plus filesView
with syncFilesEntry replacing syncPreviewRail; Workbench dual entry;
PreviewPane gains a back-to-list control; TaskSidebar project switcher;
TopBar drops project identity; Conversation drops the header stop button
and gains the input status line; workspaces.ts drops the artifacts builtin.
Refs: three-column workbench proposal (2026-10-11), DESIGN.md 6/9.3
revision follows in the Dev repo.
Why: reproduced with headless Chromium - clicking an artifact row with no
conversation selected closed the workspace overlay and nothing else appeared.
requestPreview stores tabs under currentConvId ?? '_' while the render side
required a non-null currentConvId, so the preview pane never mounted in the
empty state. The empty state was itself a trap: currentConvId was not
persisted, so every reload landed there.
Evidence: vitest 85/85 (new preview-empty-state tests for the key contract
and orphan-tab merge); browser replay on the fixed build shows the preview
pane mounting with rendered docx content from the empty state and the
conversation restoring after reload (screenshots in the Dev task card).
Impact: previewTabsKey() unifies the read/write key; entering a conversation
adopts orphan '_' tabs; the last conversation id persists and restores once
on boot; deliverable cards toast when the artifact is gone instead of
silently doing nothing.
Refs: Dev task feature/empty-state-preview-fix; diagnosis session 2026-10-10.
Why: run.Usage now carries the WP6 split fields; the badge should show
them ('+last · cumulative · duration · tools', cache hits in the
tooltip).
Evidence: npx tsc --noEmit clean; npx vitest run 82/82 green.
Impact: display only.
Refs: Agent-System-Backend#96, N3 plan WP6
Why: .main { display: flex } overrides the UA [hidden] rule, so previewExpanded left the conversation column in the row and squeezed the preview.
Evidence: vitest test/previewfit.test.ts and test/design-guard.test.ts passed (10 tests). Headless Chromium at 1512x900: expanded .main is display:none and the preview slot is 1184px with the page centered (center delta 0); at 420px and 640px the wrapper stays column-width and the page stays left-aligned.
Impact: expanded preview fills the main row; docx pages center only when the column is wider than the page. Narrow-column fit scaling is unchanged. previewfit still returns 1 when available width is <= 0.
Refs: App.tsx hidden={previewExpanded}; styles.css .main and .docx-wrapper
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The pane pinned .ui-fp to the column height, and the docx fit host then clipped overflow, so the right-hand preview could not scroll. The fullscreen DocViewer was a second shell around the same FilePreview.
Evidence: npx vitest run (17 files, 82 tests) and npx tsc --noEmit passed. Headless Chromium laid out a 1400px docx host inside a 764px pane (body scrollHeight 1400) and a pdf iframe that filled that pane.
Impact: .preview-pane-body is the scroll owner. Docx refit also watches page height. The overlay viewer is gone; expand/restore and text edit/save live on the pane.
Refs: Agent-System-Dev#196, Agent-System-Frontend#44
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Headers, HITL details, tool lists, and status words used faint text, semantic colors were used as text, and the molv sidebar primary button matched the navy surface. There was no control for --ui-font-size.
Evidence: npx vitest run (17 files, 81 tests), npx tsc --noEmit, and knip unused exports 43 <= 43 passed in modules/frontend.
Impact: Content classes move to --text-dim or *-text. Sidebar active, focus, and the new-task button use the sidebar tokens. Settings gains 小/标准/大/特大, stored only in localStorage as ui.fontSize. Before: no font-size control and a 1:1 new-task button on molv. After: four local sizes and a light sidebar button.
Refs: feature/contrast-typography
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: tsconfig.tsbuildinfo is a local tsc cache, and a fresh typecheck rewrites it, which leaves the frontend submodule dirty.
Evidence: git ls-files still listed it after the ignore rule, and git status showed it modified after npx tsc -b.
Impact: The file remains on disk and is now ignored. No source or preview behavior changes.
Refs: feature/docx-paper-preview
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: tsc -b writes tsconfig.tsbuildinfo into the worktree, which showed up as a dirty submodule during the paper-preview check.
Evidence: git status listed tsconfig.tsbuildinfo as untracked after npx tsc -b.
Impact: The artifact stays local. No runtime or preview behavior changes.
Refs: feature/docx-paper-preview
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: An A4 page is wider than the preview column, and the page background followed the app theme, so body text disappeared in dark mode.
Evidence: npm test (17 files, 81 tests) and npx tsc -b passed in modules/frontend.
Impact: Docx pages stay on --paper with their own styles and shrink to fit the column. A column wider than the page, including fullscreen, keeps them at 100%. Before: section background var(--bg) and color var(--text). After: var(--paper) and var(--paper-ink).
Refs: feature/docx-paper-preview
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: tsc -b rewrote the tracked build info after the contract mirror changed.
Evidence: make build completed and left only tsconfig.tsbuildinfo dirty.
Impact: The feature branch working tree matches the production build.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Core is the authority for the shared types, and the section shape no longer needs its own export.
Evidence: make contracts copied modules/core/ts/types.ts. Frontend knip stays at 43.
Impact: Callers still read sections on the observation object.
Refs: .task/feature-sillytavern-review.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Preview setup clears a node with innerHTML = '' before appending sanitized children, and the ratchet treated that space as a write.
Evidence: npx vitest run test/html-sinks.test.ts. Non-empty assignments still match.
Impact: The preview path can keep clearing nodes. Writing HTML strings in components still fails the test.
Refs: docs/records/2026-10-04-sillytavern-review.md
Why: Keys must be added without writing masks back, and a turn needs copy, exclude, and regenerate without deleting history.
Evidence: npx tsc -b. Session notes and candidate switches call the new conversation and message routes.
Impact: Settings gains key, lore, and slash-template panels. Excluded messages stay visible and are dimmed.
Refs: docs/records/2026-10-04-sillytavern-review.md
Why: The conversation and office fallbacks wrote marked output straight into the page, including converted docx HTML.
Evidence: html-sinks ratchet forbids new innerHTML writes outside @ui safeHtml.
Impact: Conversation keeps line-break rendering. Unsafe links are stripped after docx and pptx preview. dompurify is deduped with the UI package.
Refs: feature/sillytavern-review
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: tracked workspace writes need a visible way to return files to the state before a user message.
Evidence: the confirm lists the files the rewind API would restore or skip.
Impact: the control only calls the new checkpoints and rewind endpoints. Usage may include the model that served.
Refs: feature/durable-rewind
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Plan, default, and auto-edit have to be visible where the user sends a message, and confirmation cards need to show the arguments.
Evidence: The mode control and card detail only use existing color, radius, and type tokens.
Impact: An empty mode still renders as 默认. Confirm cards use the wait color instead of the success color.
Refs: feature/permission-hooks
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The phase type has to be consumed by the screen that branches on it, or the unused-export ratchet fails.
Evidence: knip stays at the existing baseline after the conversation annotates ShellPhase.
Impact: No behavior change.
Refs: feature/runstate-shell
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The send button, stop button, and queued sends each looked at a different slice of run state, so a reconnect or a late assistant message could disagree.
Evidence: A depth-4 walk of send, HTTP results, SSE, reconnect, conversation switch, and flush keeps busy aligned with an in-flight run, hides the live bubble once an assistant message exists, and keeps one user message per request id.
Impact: The event listener list is exhaustive over the stream contract, including run.check. A disconnected stream shows 重连中.
Refs: feature/runstate-shell
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Dead exports were unchecked, and async run state needs one owner.
Evidence: knip unused exports stay at 43. design-guard passed.
Impact: A ratchet on export count. No runtime behavior change.
Refs: feature/architecture-baseline
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The pull request cannot merge while the head is behind main.
Evidence: origin/main is 539aae6, one commit ahead of this branch.
Impact: Keeps the live-run fix and the retried request-id fix together.
Refs: Frontend #38
Why: Stamping the user message with the run id made the live bubble look settled, and the composer stayed on 发送 until the next reconcile.
Evidence: npx tsc -b passed. vitest 75, including hasSettledReply and mergeStartedRun. On a mock server the button became 停止/排队 within 250ms and the reply grew while the run was open.
Impact: Composer, live bubble, and agent tracker only. A stored user message no longer hides the stream. Reopening replays from resumeSeq when the server sends it.
Refs: Dev fix/live-run-visibility
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: A busy 409 used to enqueue a new id, so the retry created a second user message after the server had already stored the first.
Evidence: vitest run passed, 13 files and 69 tests, including the outbox retry case.
Impact: Data layer only. sendMessage and outbox flush reuse the queued message id as requestId. No screen changes.
Refs: Dev feature/pi-durable-borrow
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Deliverable cards appeared as soon as a file was registered, and the last draft was shown as the answer before the run had finished composing one.
Evidence: vitest run test/live.test.ts test/toolkind.test.ts test/messageparts.test.ts test/design-guard.test.ts
Impact: Cards wait until the run ends. A finalizing phase keeps earlier text in the process block, and check rows render there.
Refs: feature/final-reply-compose
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Streaming tool arguments looked like a frozen cursor, and a failed run folded the only narration into the collapsed process block.
Evidence: npx vitest run in modules/frontend, 13 files, 66 passed. npx tsc -b reports only a pre-existing missing docx-preview module.
Impact: Live tools move composing to running to done with arguments and output; finished runs lift the last narration; same-run errors and duplicate files collapse into the reply.
Refs: feature/stream-display-overhaul
Co-authored-by: Cursor <cursoragent@cursor.com>
weichen
