Why: The frontend contract mirror has to include script steps and critic fields before a workflow can be described to the client.
Evidence: make contracts copied modules/core/ts/types.ts. No screen or event changed.
Impact: contracts.ts only. The existing plan, inbox, and HITL cards stay the UI.
Refs: feature/agent-harness-p4
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: A background child should not look like the conversation is busy, and its completion note needs a compact row instead of a full assistant bubble.
Evidence: vitest run passed, including design-guard. tsc -b passed.
Impact: The stop button cancels the conversation run. System notices fold with the shared Collapsible.
Refs: feature/agent-harness-p3b
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Plugin pages must mount through a host bridge, and button versus chat command parity needs a check that vitest does not execute as a unit test.
Evidence: vitest run passed. The Playwright file lives under e2e and is outside the tsconfig include.
Impact: @playwright/test is not a required dependency yet.
Refs: Agent-System-Dev feature/base-extension-roadmap
Why: Plugin pages must mount through a host bridge, and button versus chat command parity needs a check that vitest does not execute as a unit test.
Evidence: npx vitest run passed. The Playwright file lives under e2e and is outside the tsconfig include.
Impact: loadPluginPage expects mount/unmount. @playwright/test is not a required dependency yet.
Refs: docs/proposals/2026-10-01-plugin-contract-v2.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Process output folded on every text delta, successful tool
results disappeared after refresh, and generated files stayed plain
text instead of a card that opens a closable preview.
Evidence: npx tsc -b --noEmit (clean); npx vitest run (51 passed).
Impact: Live work stays expanded until the run ends, then folds to
one worked-time line. File cards open a per-conversation preview pane.
docx renders page by page, with mammoth if that fails.
Refs: feature/chat-timeline-preview
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: On an empty workspace the primary button and the empty state both say to start by creating a task, but the action only toasted that a project was required.
Evidence: A fresh embedded instance returns an empty project list, so the first click never opened a conversation.
Impact: The first task now creates 「默认项目」 and opens it. Existing projects are unchanged.
Refs: walkthrough of the empty workspace on main
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: a busy conversation rejected further input, and the file list lived
only behind the workspace overlay.
Evidence: npx tsc --noEmit and vitest 42 tests, including outbox
append/edit/promote. The queue uses Collapsible and design tokens.
The files tab reuses ArtifactsPanel.
Impact: while a run is active, Enter queues the draft (edit, delete, or
send-now). Finished and failed runs drain the queue. HTTP 409 also
enqueues instead of only toasting. Backend serialization is unchanged.
Refs: task feature/input-queue
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: docx preview already sniffs ZIP magic so a text file renamed to
.docx stays readable. xlsx and pptx loaders still fed those bytes to
the parser and surfaced a raw library error.
Evidence: npx tsc --noEmit and npx vitest run (40 tests) on this branch
after merging origin/main. Legacy .xls/.ppt skip the sniff because they
are not OOXML.
Impact: pseudo xlsx/pptx render as text with a short note. Real ZIP
workbooks and decks still use the existing parsers.
Refs: Dev PR #96, task feature/preview-sniff-fallback
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: field report two — the final answer collapsed into the process
area (the CF0 long-reply fold swallowed 2000+ char replies and the
model inlines narration, so users saw "everything folded") and the
artifacts workspace panel preview clicks did nothing (the full-screen
overlay stayed up, hiding the rail the preview opened in).
Evidence:
- Conversation.tsx: removed the CF0 LongReplyCollapse entirely — the
final output contract is "the reply to the user is always fully
visible; only process parts fold"; deliverable cards stay pinned
after the reply
- Workspaces.tsx ArtifactsPanel: row click and fullscreen action now
close the workspace overlay before requesting the rail/fullscreen
preview, so the preview is actually on screen
- vitest 40/40, tsc clean; no design-guard regressions (no new colors,
keyframes, or inline font sizes)
Impact: replies are readable in place; ⌘1 panel clicks visibly open
previews; no other surfaces changed.
Refs: task feature/agent-native-deliverables-r1 (backend: artifact
backfill root fix + agent-native deliverables)
Why: the pre-commit/push mimosa hooks write .mimosa/hook-state inside
the work tree; untracked it marks the Dev superproject dirty
(precheck SHARED_WORKSPACE_BUSY) and blocks card-close gates.
Evidence: git status showed ?? .mimosa after commits; with the ignore
the superproject returns clean. No tracked files affected.
Impact: none at runtime; cache stays local.
Refs: task feature/deliverable-docx-r1 follow-up
Why: tool artifacts were only reachable via the collapsed process
section or the artifacts workspace; users had to read model text to
learn a file existed. Deliverables should appear directly in the
conversation as an openable card (ZCode/Kimi pattern) that opens the
right-rail preview.
Evidence:
- Conversation.tsx artifact_ref renderer upgraded from a small chip to
a deliverable card (icon + name + kind/size meta + preview affordance
+ single download action); whole card click = rail preview (the one
preview placement, fullscreen still reachable from the rail header)
- live.ts projects artifact items as content parts pinned after the
reply text, isomorphic with the new llm.go persisted order; card is
visible while streaming
- styles: deliverable-card uses tokens only (radius-lg card, bg-raised,
hairline border, mk-rise-in entrance); design-guard suite green
- vitest 40/40 (live.test pinned-tail case updated, messageparts.test
covers new persisted order plus legacy inline order); tsc clean
Impact: every generated artifact (docx/pdf/image/…) now surfaces as an
in-chat card immediately, click-to-preview in the right rail with real
formatting for styled docx; workspace panel (⌘1) unchanged as the
file-system fallback.
Refs: task feature/deliverable-docx-r1 (backend: write_docx + pinned
artifact_ref)
Why: Keep the task branch on the shared baseline before acceptance.
The merged CF3 code carried a bare hex fallback (#d8d8d8) and off-scale
font size in .inbox-scope — the design guard caught it; fixed to tokens
(--border / --radius-full / --fs-ui-sm / --bg-hover) per DESIGN.md.
Evidence: clean ort merge; npx tsc -b clean; npm test -> 39 passed
(guard green after the fix); npm run build succeeds.
Impact: brings the project inbox dual view into the branch with its
styles brought onto the token system; no behavior changes.
Refs: task feature/ui-tokens-v3.
Why: Approved redesign phase 4: process folds, long-reply folds, thinking
blocks and context-observation fold move to the @ui Collapsible primitive
(two-layer height+fade animation, delayed unmount, open-state memory);
tool calls render as ToolBlock rows with running sweep and paired results;
the live bar gains the ZCode worked-time line with sweep; input placeholder
states the backend's honest busy semantics (serialized per conversation).
Evidence: npx tsc -b clean; npm test -> 39 passed (details-guard included);
npm run build succeeds; combined make test recorded in the Dev task card.
Impact: Conversation.tsx replaces all <details> usage (guard-enforced);
ProcessParts pairs tool_call with its trailing tool_result into ToolBlock;
styles.css rewrites fold selectors for the Collapsible structure and adds
message mount animation (mk-rise-in); no contract or API changes.
Refs: Agent-System-Dev task feature/ui-tokens-v3 (phase 4).
Why: D49/D51 — aggregated project view (project cards + other
conversations' pending) so todos survive conversation switches;
D50 — subscribe the project channel via scope=project.
Evidence: api endpoints (listProjectCards/cardToProject, card.* named
SSE events — also fixes live card events previously dropped);
store projectCards/inboxScope/applyProjectEvent/cardToProject with
badge combining both scopes; Workbench scope tabs + grouped ProjectInbox;
styles for scope control.
Impact: existing conversation inbox behavior unchanged; badge now counts
both scopes.
Refs: documents proposal 2026-09-30-card-stream-artifact-unify-project-inbox.md (D48-D51).
Why: Approved redesign phase 3: one artifact = one preview placement
(right rail by default; full-screen overlay as the only large-window form)
plus a single download button per surface. Removes the redundant
three-action chips/rows and the nested modal chrome in AgentTracker.
Evidence: npx tsc -b clean; npm test -> 38 passed. Combined runs recorded
in the Dev task card.
Impact: new store.requestPreview() is the single preview entry (Command
semantics; placement derived there); artifact chips and workspace rows
reduce to title-click = rail preview, eye = full preview, download;
docBridge card opens route through the same entry; AgentTracker drops its
inner .modal shell (OverlayHost is the only chrome). No API changes.
Refs: Agent-System-Dev task feature/ui-tokens-v3 (phase 3).
Why: Approved redesign phase 2 (IA): normal users run projects entirely
through conversation; system capabilities (user management, agent profiles,
plugin management, plugin-declared workspaces) belong behind a dedicated
admin entry, not in the user-facing sidebar or settings. The sidebar
workspace strip is removed; artifacts stay reachable via the right-rail
tab and conversation chips.
Evidence: npx tsc -b clean; npm test -> 38 passed (design-guard ratchet
still green). Full combined run recorded in the Dev task card.
Impact: new AdminZone full-screen view with a declarative section registry
(users/profiles/plugins/plugin-workspaces via UIRenderer); sections moved
out of SettingsModal (now personal preferences only) and AdminUsersModal
(now password change only); store ui.adminUsersOpen renamed adminZoneOpen;
workspaceEntries narrowed by role so plugin entries no longer take Cmd1-9
for regular users; theme toggle moved into the account menu.
Refs: Agent-System-Dev task feature/ui-tokens-v3 (phase 2).
Why: Phase 1 of the approved frontend redesign: re-skin the shell to the
Zai look (floating rounded panels on an inset base with 4px gutters),
invert primary buttons to the mono scheme, and move every font size onto
the --fs-ui-* scale. Also wires @ui/motion.css and replaces local
keyframes with mk-* utilities.
Evidence: npx tsc -b clean; npm test -> 38 passed (new host-side
design-guard ratchet); npm run build succeeds. Combined Dev make test
recorded in the Dev task card.
Impact: styles.css fully re-tokenized (panel borders/radius, primary
button color-mix hover, login brand panel in near-mono navy); inline
fontSize literals in components replaced with scale tokens except one
decorative glyph in the ratchet baseline. No component class names or
behavior changed.
Refs: Agent-System-Dev task feature/ui-tokens-v3; DESIGN.md (parent repo).
Why: agent-generated ".docx" artifacts were historically plain text bytes
written via string-only write tools; the docx preview loader fed them
straight to mammoth/JSZip, which threw "Can't find end of central directory"
on the meeting-notes docx from the review session.
Evidence: docxLoader now checks the PK\x03\x04 (or empty-archive PK\x05\x06)
magic before mammoth; non-ZIP bytes render as markdown with a readable note
instead of an error. vitest 33/33 pass, tsc --noEmit clean.
Impact: stock fake-docx artifacts become readable immediately; real docx
previews unchanged. Rendering only, no API/contract change.
Refs: .task/feature-preview-sniff-fallback.md
Why: the adjudicated plan (D1-D4) calls for office/link previews with
loaders injected from the host, right-rail light preview alongside the
overlay viewer, and artifact deletion kept as an API capability with no
UI entry.
Evidence: previewsetup.ts injects docx (mammoth), xlsx (SheetJS, sheet
tabs, 200-row cap), pptx (pptx-preview full-fidelity; pptx2html was
dropped for a broken module entry — tXml case mismatch breaks Linux
builds — per the D2 fallback clause), and the /api/url-preview fetcher,
all via dynamic import: the main entry grows ~12KB while xlsx (430KB),
pptx (1.2MB) and jszip (98KB) load on first preview. DocViewer becomes a
FilePreview container (keeps text-family editing); Workbench gains a
preview tab with rail-preview state (store.railPreview); artifact chips
and panel rows gain a rail-preview action; the panel delete button is
removed while api.deleteArtifact and DELETE /api/artifacts/{id} remain.
tsc clean; vitest 33 passed; production build succeeds.
Impact: previews work end to end for the whole format matrix; deletion
stays plugin-reachable only.
Refs: task feature/file-preview-caps; ui fb76239; core 4d29301; backend
6d4ccfd.
Why: GHSA-82fw-gwwq-j7x9 (path traversal / arbitrary file read via @vitest/mocker redirect mock) affects vitest 2.1.0-4.1.10; the locked 3.x toolchain was inside the range. Dev-only dependency, but the fix is a routine bump — follow-up to the 2026-09-30 dependency CVE inventory.
Evidence: vitest and @vitest/mocker now at 4.1.11 (first fixed release); npm audit reports 0 vulnerabilities; npx vitest run 33/33 green on vitest 4.
Impact: package.json/package-lock.json devDependency bump only; no production dependency or source change.
Refs: Dev-side task fix/dep-vuln-bumps (gitlink pin follows after combined verification)
Why: agents rediscovered Gitea hosts and credential locations from
scratch each session, and the retired git-panel host kept failing with
TLS broken-pipe errors before the right host was recalled.
Evidence: curl git-home...:30319/api/v1/version with token from
~/.secrets/gitea -> {"version":"1.27.3"}; ssh -p 31411 git@git-ssh...
(BatchMode) -> "Hi there, weichen!" with key ~/.ssh/home_gitea (ED25519,
SHA256:GYXaOXIMeloQ1fKSfdJ/ndgPQzs/Ix7Hk62ekfkfQ20). Same section already
merged in Agent-System-Dev (PR #73).
Impact: AGENTS.md only; records hosts and credential locations, no
secret values (baseline hard boundary 2).
Refs: Agent-System-Dev PR #73
Why: the merge commit left tsc broken in two spots - an artifact.created
handler still pushing to the removed LiveState.parts, and main's newer ui
icons not yet checked out in this worktree; also Part was imported unused.
Evidence: artifact.created now appends an artifact LiveItem (process=true,
matching the backend's ae8e5d0) projected to artifact_ref by
liveToMessage; new test asserts it lands after its tool call. tsc -b exit
0 after initializing the ui submodule checkout; vitest 33/33.
Impact: live artifact chips render inside the process stream and fold on
reconcile; no functional change elsewhere.
Refs: feature/chat-process-split-r1
Why: main advanced with PR #15/#16 (part-order preservation, artifact
downloads, cross-session UI); task-sync requires the branch to carry it.
Evidence: resolution keeps the arrival-ordered items stream from this
branch (tool-driven process marking, pure/testable logic in live.ts) and
adopts main's rendering and ids: groupParts in-place process blocks,
real tool callIds on live tool parts, messageparts utilities, and the
restored phase-driven showProcessInline predicate alongside the CF0
long-reply excerpt that main's rewrite had dropped. vitest 32/32
(includes main's messageparts suite and this branch's 12 live cases);
tsc -b clean.
Impact: single coherent live pipeline - items accumulate in event order,
liveToMessage projects Parts with tool-driven process marks, rendering
folds consecutive process blocks in place; responding still collapses
the process area (Cursor-style) as originally reviewed.
Refs: feature/chat-process-split-r1
Why: liveToMessage emitted all text buckets then all tools, while the
persisted message interleaves per turn (thinking, text, calls), so parts
reordered on reconcile with 2+ tool turns. Commentary stayed in the reply
area until the next turn's text bucket appeared, because the process flag
was positional (i < length - 1) instead of tool-driven.
Evidence: LiveState now holds an arrival-ordered items stream
(thinking/text deltas merge into their last segment; tools land in place),
so event order - which matches the backend's per-turn append order - maps
1:1 onto persisted parts. A text segment is process iff a tool.call lands
before the next thinking/text segment, aligning with the backend's
Process: len(calls) > 0; the collapse predicate moved to testable
showProcessInline. Long replies (>800 chars) now show a 100-char excerpt
and label the count as Markdown source length. vitest 27/27 pass (was 5
live cases, now 12 incl. multi-turn order, immediate process marking,
collapse predicate); tsc -b clean.
Impact: streaming UI no longer reshuffles parts across turns or on final
reconcile; tool-turn commentary folds into the process area as soon as the
tool call arrives. Residual edge (documented in live.ts): a silent
tool-only turn with no text and no thinking attributes its call to the
previous text segment.
Refs: code review of feature/chat-process-split-r1 (2026-09-30)
Why: 642299d committed node_modules as a symlink pointing at the absolute
path /home/weichen/projects/Agent-System-Dev/modules/frontend/node_modules.
The .gitignore rule "node_modules/" only matches directories, not symlinks,
so the link slipped through and other clones would receive this machine's
path while normal installs get bypassed.
Evidence: git ls-tree HEAD node_modules => mode 120000 blob before this
commit; git check-ignore -v node_modules exits 0 with rule
".gitignore:2:node_modules" after unstaging the link.
Impact: fresh clones no longer carry a machine-specific symlink; local
installs work normally. No code changes.
Refs: code review of feature/chat-process-split-r1 (2026-09-30)
weichen
