Agent live card now uses the sentinel pattern to detect when it scrolls
out of view. When stuck, it collapses to a compact header bar with brand
styling and backdrop blur, with a ChevronUp button to scroll back.
When scrolled back into view, the card seamlessly expands to full view.
Also adds semantic colors to Sonner toast icons (success/info/warning/
error/loading) and fixes icon-to-text alignment in toasts globally.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When a member-started thread root @mentions the assignee agent, replies
in that thread should trigger on_comment — the thread is a conversation
with the agent, not a member-to-member chat.
Previously isReplyToMemberThread only checked the reply content for
assignee mentions. Now it also checks the parent (thread root) content.
This fixes a gap where path 1 (on_comment) suppressed the trigger and
path 2 (on_mention) skipped the assignee, leaving no trigger path.
* feat(agents): hide archived agents from default list
Archived agents are now filtered out of the default agent list view.
A toggle button (archive icon) appears when archived agents exist,
allowing users to switch between viewing active and archived agents.
The @mention suggestion list already filters out archived agents.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(agents): show "No active agents" when all agents are archived
When there are archived agents but no active ones, the empty state now
shows "No active agents" instead of "No agents yet" to avoid confusion.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Devv <devv@Devvs-Mac-mini.local>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The useEffect that scrolls to and highlights a comment had timeline.length
in its dependency array. When a new reply was posted, timeline.length changed,
re-triggering the scroll and highlight animation. Added a ref to track whether
we've already highlighted for the current highlightCommentId so it only fires once.
When a top-level comment @mentions an agent (non-assignee), subsequent
replies in the same thread now also trigger that agent via on_mention.
Previously only the current comment's mentions were checked, so replies
without an explicit re-mention would silently skip the agent.
Extends enqueueMentionedAgentTasks to accept the parent comment and
merge its parsed mentions (deduplicated) into the trigger set, reusing
all existing guards (self-trigger, assignee skip, visibility, dedup).
Closes MUL-177
Add architecture comments to content-editor.tsx, markdown-paste.ts,
extensions/index.ts, mention-view.tsx, content-editor.css, and
preprocess.ts explaining: why single markdown pipeline, why
data-pm-slice for paste detection, typography benchmarks, mention
card sizing rationale, and what was removed from the old system.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Markdown paste: replace heuristic-based detection with a single
deterministic check — only use HTML clipboard path when source is
another ProseMirror editor (identified by data-pm-slice attribute).
All other pastes (VS Code, text editors, terminals, .md files) parse
text/plain as Markdown via @tiptap/markdown.
Inline code: add box-decoration-break: clone and line-height: 2 so
multi-line inline code renders with proper spacing between lines.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Use var(--brand) instead of var(--primary) for link color (blue vs
near-black)
- Add default underline at 40% opacity, full opacity on hover
- Remove Tailwind HTMLAttributes from Link extensions — let CSS control
all link styling uniformly
- Mention cards unaffected (a.issue-mention overrides with color: inherit)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Restore card to readable size (py-0.5, px-2, text-xs, rounded-md)
- Add max-w-72 and truncate on title for long issue names
- Move vertical-align: middle to [data-node-view-wrapper] (outermost
inline element) instead of inner <a> — fixes centering within line
- Always render card style even when issue not in store (fallback shows
identifier only)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Reduce card dimensions so it doesn't overflow the paragraph line box:
- Padding: py-0.5 → py-px (4px → 1px vertical)
- Font size: text-sm → text-xs with leading-relaxed
- Alignment: vertical-align: middle via CSS selector
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace three divergent data paths (Marked HTML loading, regex post-processing
saving, separate paste parsing) with one symmetric path through @tiptap/markdown.
Key changes:
- Create features/editor/ module with ContentEditor (unified edit+readonly)
and TitleEditor, replacing components/common/ editor files
- Load content via contentType: 'markdown' instead of markdownToHtml() hack
- Save content via editor.getMarkdown() directly, no post-processing
- Merge RichTextEditor + ReadonlyEditor into single ContentEditor with
editable prop
- Extract extensions into separate modules (mention, file-upload,
markdown-paste, submit-shortcut, code-block-view)
- Extract shared preprocessMentionShortcodes to components/markdown/mentions.ts
- Add copyMarkdown utility for clipboard operations
- Upgrade all @tiptap packages from 3.20.5 to 3.22.1 (lexer isolation fix,
HTML entity roundtrip fix, table alignment support)
- Delete markdownToHtml.ts, readonly-editor.tsx, and 10 old component files
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(agent): instruct agents to use download_url for attachments
Agents were not aware of the signed vs unsigned URL distinction in
attachments, causing failures when trying to read images. Added an
Attachments section to the generated CLAUDE.md/AGENTS.md template that
tells agents to always use `download_url`. Also increased signed URL
expiry from 5 to 30 minutes to better accommodate agent processing time.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(cli): add `multica attachment download` command
Adds a dedicated CLI command for downloading attachments by ID. The
command fetches attachment metadata from the API (which returns a fresh
signed URL), downloads the file, and saves it locally. This eliminates
the need for agents to understand signed vs unsigned URLs.
Changes:
- New `multica attachment download <id>` CLI command
- New `GET /api/attachments/{id}` backend endpoint
- `DownloadFile` helper on APIClient
- Updated CLAUDE.md template to document the command
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): sanitize filename and add download size limit
- Use filepath.Base on attachment filename to prevent path traversal
- Add 100MB size limit to DownloadFile (matches upload limit)
- Include response body in download error messages for debugging
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Devv <devv@Devvs-Mac-mini.local>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Heading hierarchy: H1 bumped to 1.125rem with letter-spacing to
distinguish from H2 (both were 1rem). Margins normalized to 0.5rem
baseline rhythm.
- List items: increased spacing from 0.125rem to 0.25rem for readability.
Remove paragraph margins inside list items (Tiptap wraps li content
in <p> tags which inherited 0.5rem margins).
- Nested lists: bullet style progression (disc → circle → square) and
numbering progression (decimal → lower-alpha → lower-roman).
- Blockquotes: tighter paragraph spacing inside, nested blockquotes get
lighter border for depth indication.
- Inline code: border-radius uses semantic --radius-sm token.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace @tiptap/markdown's beta contentType: "markdown" parser with a
dedicated marked-based HTML pipeline for loading markdown content.
The @tiptap/markdown parser silently drops content in complex documents
(tables, nested lists, mentions). Instead, we now:
1. Pre-convert mention links to <span data-type="mention"> HTML
2. Render markdown to HTML via a dedicated Marked instance with a custom
renderer that wraps table cell content in <p> tags (required by
Tiptap's TableCell block+ content spec)
3. Load as HTML — Tiptap's ProseMirror HTML parser handles everything
4. Keep @tiptap/markdown extension only for getMarkdown() serialization
Also adds Table extension support and aligns CSS with the old Markdown
component's minimal mode styling.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Use useState for selectedKey instead of deriving directly from
useSearchParams(), so the issue ID from ?issue=<id> is reliably
captured on mount. window.history.replaceState() doesn't always
sync back to useSearchParams() in Next.js, causing the detail
panel to show empty when entering via a shared inbox link.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The integration tests hardcoded the old default JWT secret while .env
sets a different JWT_SECRET, causing all authenticated requests to fail
with 401. Use auth.JWTSecret() so tests stay in sync with the server.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Paste/drop and attachment button previously used separate upload paths.
The button uploaded first then called insertFile (which replaced the
current selection), while paste inserted a blob preview first. This
caused the second image to overwrite the first when both were used.
Now both paths share the same flow via uploadAndInsertFile():
blob preview with uploading animation → background upload → replace URL.
- Extract shared uploadAndInsertFile() function
- Replace insertFile ref method with uploadFile (inserts at doc end)
- Simplify FileUploadButton to onSelect(file) — no more onUpload/onInsert
- Wire onUploadFile in comment edit mode (was missing, upload was no-op)
- Unify image border-radius CSS for both editing and readonly modes
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(daemon): add opencode as supported agent provider
Add opencode backend alongside claude and codex. The backend spawns
`opencode run --format json`, parses streaming JSON events (text,
tool_use, error, step_start/finish), and supports --prompt for system
prompts. Includes CLI detection, AGENTS.md runtime config, native skill
discovery via .config/opencode/skills/, and 21 tests covering handlers,
JSON parsing, and integration-level processEvents scenarios.
* chore: add .tool-versions to gitignore
When viewing an issue in the inbox, WS events like issue:updated and
inbox:new triggered full store refetches, causing unnecessary loading
flashes and redundant API calls. Now these events update the store
in-place using the event payload data instead of refetching everything.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(agent): replace hard delete with archive/restore
Replace agent deletion with soft archive pattern. Archived agents
are preserved in the database with all historical references intact
but cannot be assigned, mentioned, or trigger tasks.
Backend:
- Add archived_at/archived_by columns to agent table (migration 031)
- Replace DELETE /api/agents/{id} with POST /api/agents/{id}/archive
- Add POST /api/agents/{id}/restore endpoint
- ListAgents excludes archived by default (?include_archived=true to include)
- Skip archived agents in task triggers (on_assign, on_comment, on_mention)
- Block assignment to archived agents
- Cancel pending tasks on archive
- New events: agent:archived, agent:restored (replacing agent:deleted)
Frontend:
- Agent type includes archived_at/archived_by fields
- Mention autocomplete and assignee picker filter out archived agents
- Agent list shows archived agents with muted styling
- Agent detail shows archive banner with restore button
- Delete button replaced with Archive button and updated confirmation dialog
- API client: archiveAgent/restoreAgent replace deleteAgent
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(agent): self-review fixes for archive feature
- Fix: workspace store now fetches agents with include_archived=true
so archived agents are actually visible in the frontend (the archived
UI was dead code before — ListAgents excludes archived by default)
- Fix: add error logging for CancelAgentTasksByAgent in ArchiveAgent
- Fix: add idempotency guards — return 409 Conflict when archiving
an already-archived agent or restoring a non-archived agent
- Fix: revert unnecessary extra GetAgent query in ReconcileAgentStatus
(archived agents won't have running tasks after CancelAgentTasksByAgent)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Override Link extension `inclusive: false` via `.extend()` to decouple
it from `autolink: true`. Tiptap's source ties `inclusive` to `autolink`,
causing typed text after a link to inherit the link mark.
Also set `linkOnPaste: false` — autolink's PasteRule still auto-detects
pasted URLs without the sticky cursor issue.
Refs: ueberdosis/tiptap#2571, ueberdosis/tiptap#4249
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Align the CI backend job with the Go version declared in server/go.mod
and used in the Dockerfile (golang:1.26-alpine).
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add an end-to-end onboarding guide covering login, daemon setup,
runtime verification, agent creation, and first task assignment.
Updated both English and Chinese READMEs.
* feat(daemon): support direct download update for non-Homebrew installs
Previously, CLI auto-update only worked for Homebrew installations. Non-brew
binaries would fail with "not installed via Homebrew". Now the daemon and
`multica update` fall back to downloading the release binary directly from
GitHub Releases when Homebrew is not detected.
Also fixes:
- Daemon restart now uses the current executable's absolute path instead of
searching PATH, ensuring the updated binary is used
- Brew installs preserve the symlink path so the new Cellar version is picked up
- Daemon startup logs now include the CLI version
- Update UI auto-clears "restarting" status after 5s to show the new version
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(cli): remove dead DetectNewBinaryPath and guard against nil latest version
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The runtime detail page was showing the agent CLI version (claude/codex)
as "CLI Version" because metadata.version stored the agent version from
agent.DetectVersion(). The multica CLI version was never sent.
Fix: daemon now sends cli_version in the registration request, server
stores it as metadata.cli_version alongside the existing agent version,
and frontend reads metadata.cli_version.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Align website title, description, and meta tags with the landing page
messaging. Add Open Graph, Twitter Card tags, sitemap.ts, and robots.ts.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(runtime): support CLI update from web runtime page
Add the ability to update the CLI daemon from the web Runtime detail page.
When a newer version is available on GitHub Releases, an update button
appears. Clicking it sends an update command through the server to the
daemon via the heartbeat mechanism (same pattern as ping). The daemon
executes `brew upgrade`, reports the result, and restarts itself with the
new binary.
Changes across all three layers:
- Frontend: version display, GitHub latest check, UpdateSection component
- Server: UpdateStore (in-memory), heartbeat extension, 3 new endpoints
- CLI: shared update logic, daemon handleUpdate + graceful restart
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(runtime): handle 'running' status in ReportUpdateResult
The daemon sends {"status":"running"} when it starts executing the
update, but ReportUpdateResult treated any non-"completed" status as
failure — immediately marking the update as failed before brew upgrade
even ran.
Fix: use a switch statement to handle "running" as a no-op (status is
already "running" from PopPending), and also timeout running updates
after 120 seconds in case brew upgrade hangs.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add NOT EXISTS check to ClaimAgentTask SQL to prevent claiming a queued
task when the same issue already has a dispatched/running task. This
ensures serial execution within an issue while preserving parallel
execution across different issues (concurrency group pattern).
Also add defensive guard in the frontend task:dispatch handler to avoid
replacing an active task's LiveLog timeline mid-execution.
Closes MUL-183
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(upload): remove file type allowlist to support all file types
Removes the hardcoded MIME type allowlist from both frontend and backend
that was blocking uploads of file types like Word documents (.docx).
File size limit (10 MB) is still enforced. Content type detection is
preserved for metadata storage.
Closes MUL-123
* feat(upload): increase file size limit from 10 MB to 100 MB
Updates both frontend and backend to allow uploads up to 100 MB.
* feat(mentions): support @mentioning issues in comments
- Extend MentionItem type to include "issue" alongside "member"/"agent"
- Add issue search (by identifier and title) to mention suggestion dropdown
- Render issue mentions with CircleDot icon in autocomplete popup
- Issue mentions serialize as [MUL-117 Title](mention://issue/id) (no @ prefix)
- Markdown renderer shows issue mentions as clickable links to /issues/:id
- Backend mentionRe regex updated to match issue mention type
* feat(mentions): auto-expand issue identifiers and add mention format to agent instructions
1. Path A — CLAUDE.md template (runtime_config.go):
Add a "## Mentions" section teaching agents the mention serialization
format for issues, members, and agents. All agents automatically
receive this via the auto-generated CLAUDE.md.
2. Approach 2 — Server-side auto-conversion (internal/mention/):
New ExpandIssueIdentifiers() utility that scans comment content for
bare issue identifiers (e.g. MUL-117) and replaces them with
[MUL-117](mention://issue/<uuid>) mention links. Skips code blocks,
inline code, and existing markdown links. Integrated into both:
- handler.CreateComment (HTTP API path)
- service.createAgentComment (agent task output path)
When clicking an inbox notification, the issue detail now scrolls to and
briefly highlights the relevant comment. Also adds a floating "Jump to
bottom" button on issue pages with long timelines.
Backend: store comment_id in inbox notification details for new_comment
and reaction_added events. Frontend: pass highlightCommentId through to
IssueDetail, add id attributes to comment elements, and track scroll
position for the jump-to-bottom button.
- ListAgents: private agents are now visible to all workspace members
(previously hidden from non-owner members)
- Mentions: private agents can only be @mentioned by the agent owner or
workspace admin/owner; regular members' mentions of private agents are
silently ignored
- Settings (update/delete/skills) and assign were already correctly
restricted in previous PRs
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
