The bare cache used a mirror-style fetch refspec
(+refs/heads/*:refs/heads/*) which collided with worktree-locked
refs/heads/agent/<task> branches once those branches were pushed
back to origin as PRs. git fetch aborted with "refusing to fetch
into branch ... checked out at ...", the error was swallowed as a
warning, and every subsequent checkout reused the snapshot from
the original clone.
Fix:
- Clone / migrate bare caches to a remote-tracking layout
(+refs/heads/*:refs/remotes/origin/*) so fetched heads never
land in refs/heads/*.
- Resolve the base ref from refs/remotes/origin/HEAD with a
5-level fallback (verified origin/HEAD symref to origin/main
or origin/master to the bare HEAD bridged into origin/<same>
to single-entry origin/* scan to bare HEAD for legacy caches).
- Refuse to guess when refs/remotes/origin/* has multiple
candidates and none match a known fallback, so CreateWorktree
fails loudly instead of basing work on an arbitrary branch.
- Refresh refs/remotes/origin/HEAD after every successful fetch,
not just on the legacy migration path, so a cache that was
already modern picks up an upstream default-branch change.
- Verify the primary symref target actually exists so a phantom
refs/remotes/origin/HEAD from a broken set-head does not
surface a deleted branch.
- Detect legacy caches on the fly and rewrite refspec +
refs/remotes/origin/* + refs/remotes/origin/HEAD in place so
existing clones self-heal on next use.
- Serialize per-bare-repo mutation (both Sync and CreateWorktree)
with sync.Map-backed mutexes so concurrent fetch and worktree
add on the same repo cannot race on git's own lockfiles.
- Narrow the already-exists retry to actual branch-collision
errors so a path-collision no longer silently leaks a branch
into the bare repo.
Codex doesn't expose token usage through its JSON-RPC app-server
protocol. The turn/completed and task_complete notifications don't
contain usage fields.
Fix: after Codex execution finishes, scan the on-disk session JSONL
files (~/.codex/sessions/YYYY/MM/DD/*.jsonl) for token_count events.
Only files modified after the task's start time are scanned, avoiding
counting unrelated sessions. This matches the same data format the
existing runtime_usage scanner reads.
CI uses pgvector/pgvector:pg17 which doesn't ship pg_bigm. Wrap
CREATE EXTENSION and index creation in DO/EXCEPTION blocks so the
migration succeeds without pg_bigm — indexes are skipped and search
falls back to plain LIKE scans.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add owner_id to agent_runtime table to track who registered each runtime.
Backend: new delete endpoint with role-based permissions (owner/admin can
delete any, members only their own), list filtering by owner (?owner=me),
and agent dependency check before deletion.
Frontend: Mine/All filter toggle in runtime list, owner display in list
items and detail view, delete button with AlertDialog confirmation.
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(search): implement full-text search for issues
Add pg_bigm-based full-text search across issue titles and descriptions,
with API endpoint, CLI subcommand, and web Cmd+K search dialog.
- Migration 032: pg_bigm extension + GIN indexes on title/description
- Server: GET /api/issues/search?q=... with pagination and total count
- CLI: `multica issue search <query>` with table/json output
- Web: Cmd+K command palette using cmdk, with debounced search
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(search): address review feedback on search implementation
1. Escape LIKE special characters (%, _, \) in handler to prevent
matching anomalies from user input.
2. Wire AbortController signal into searchIssues fetch so in-flight
requests are actually cancelled on new input.
3. Fix offset=0 falsy check — use !== undefined instead of truthiness.
4. Merge results + count into single query using COUNT(*) OVER()
window function, eliminating the duplicate DB round-trip.
5. Exclude done/cancelled issues by default; add include_closed
parameter to API, CLI (--include-closed), and web client.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(search): default web search to include all statuses
Pass include_closed: true in the web Cmd+K search so results include
done and cancelled issues by default, matching the reviewer's request.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(search): add comment search with snippet extraction
Extend search to cover issue comments in addition to title/description.
Results are deduplicated at the issue level, with match_source and
matched_snippet fields indicating where and what matched.
- Migration 033: pg_bigm GIN index on comment.content
- SQL: EXISTS subquery for comment matching, correlated subquery for
snippet extraction, 3-tier ranking (title > description > comment)
- Server: SearchIssueResponse with match_source and matched_snippet
- Web: show comment icon + snippet below issue title when matched
- CLI: MATCH column shows source and truncated snippet
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(search): redesign search dialog to match Linear's spacious style
- Widen dialog from sm (384px) to xl (576px) with top-20% positioning
- Larger search input with icon, generous padding, and ESC hint
- Use cmdk primitives directly for full style control
- Taller result list (400px / 50vh), spacious result items (py-2.5)
- Rounded-lg items with accent highlight on selection
- Cleaner border separator between input and results
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(issues): add sub-issue support
- Backend: Add ListChildIssues SQL query, add parent_issue_id to UpdateIssue,
add GET /api/issues/{id}/children endpoint
- Frontend: Display parent issue breadcrumb and link in issue detail sidebar,
show child issues list with status icons, add "Create sub-issue" action in
dropdown menu and sidebar, pass parent_issue_id through create issue modal
- Update test mocks for new API method
* fix(issues): add parent validation, cycle detection, and improve child refresh
- CreateIssue: validate parent issue exists in the same workspace
- UpdateIssue: validate parent exists, prevent self-referencing, detect
circular parent chains (up to 10 levels deep)
- Frontend: derive child issues from store when available instead of
refetching on every global issue count change
* fix(daemon): update existing worktree to latest remote on reuse
When an agent receives a new task on the same issue, the execution
environment is reused and the repo worktree already exists on disk.
Previously, `multica repo checkout` would fail because `git worktree add`
cannot create a path that already exists — so the agent worked on stale
code from the prior task.
Now `CreateWorktree` detects existing worktrees and updates them:
fetch origin, reset working tree, then checkout a new branch from the
latest remote default branch. The previous task's branch is preserved.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(daemon): propagate actual branch name and use correct ref in worktree reuse
- Return (string, error) from updateExistingWorktree so collision-retried
branch name propagates to WorktreeResult
- Use baseRef directly instead of origin/baseRef — bare clone refspec maps
remote branches to local refs, so remote-tracking refs may not exist
- Remove redundant fetch (worktree shares object store with bare clone)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1. Separate ReportTaskUsage endpoint (POST /api/daemon/tasks/{id}/usage)
so usage is captured independently of complete/fail — fixes usage loss
for failed/blocked tasks.
2. Add usage tracking for all four providers:
- Claude: already done (stream-json message.usage)
- OpenCode: extract from step_finish.part.tokens
- OpenClaw: extract from step_end.data token fields
- Codex: extract from turn/completed and task_complete usage fields
3. Remove usage from CompleteTask payload — all usage goes through the
dedicated endpoint now.
Extract token usage from Claude Code's stream-json output in real-time
during task execution, replacing the inaccurate global JSONL log scanner.
- New `task_usage` table: tracks (task_id, provider, model) level usage
- Agent SDK: parse `message.usage` from assistant messages, accumulate
per-model and return in Result
- Daemon: convert agent usage to entries, send with CompleteTask
- Server: store usage on task completion, expose workspace-level
aggregation APIs (GET /api/usage/daily, GET /api/usage/summary)
* refactor: remove agent triggers config field
Remove the triggers field from agent configuration. The on_assign,
on_comment, and on_mention behaviors are now always enabled (hardcoded),
as decided in the Agentflow design discussion (MUL-372).
Changes:
- Database: migration 032 drops triggers column from agent table
- Backend: remove triggers from create/update agent APIs and response
- Backend: simplify trigger-checking logic to always-enabled
- Frontend: remove TriggersTab UI and AgentTrigger types
- Tests: remove trigger config unit tests (no longer configurable)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor: also remove agent tools config field
Remove the tools field from agent configuration alongside triggers.
The tools field was a placeholder — stored in the DB and shown in the
UI but never passed to the daemon or used at runtime.
- Database: migration 032 now also drops tools column
- Backend: remove tools from create/update agent APIs and response
- Frontend: remove ToolsTab UI, AgentTool type, and tools tab
- Update landing page copy
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(test): remove tools/triggers columns from test fixtures
The test fixtures still referenced the dropped tools and triggers
columns when inserting agent rows, causing CI failures.
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>
- Relax ClaimAgentTask SQL constraint from per-issue to per-(issue, agent)
serialization, allowing different agents to run in parallel on the same issue
- Update GetActiveTaskForIssue API to return all active tasks (array) instead of
just the first one
- Refactor AgentLiveCard to render one card per active task, routing WebSocket
messages by task_id for independent timelines
- Fix shouldEnqueueOnComment to use per-agent dedup so a mentioned agent's
pending task doesn't block the assigned agent's on_comment trigger
Closes MUL-160
- Add ListOpenIssues SQL query (excludes done/cancelled, no LIMIT)
- Add CountIssues SQL query for true total count
- Backend: support open_only=true param, fix total to return real count
- Frontend: two-phase fetch in issue store (all open + first 50 closed)
- Add fetchMoreClosed action for paginated closed issue loading
- Replace all hardcoded limit:200 with store.fetch() calls
Resolves MUL-369
Co-authored-by: Devv <devv@Devvs-Mac-mini.local>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Support Google login that links to existing accounts by email.
When a user who registered via email OTP signs in with Google using
the same email, they are linked to the same account.
Backend:
- Add POST /auth/google endpoint that exchanges Google auth code for
tokens, fetches user profile, and calls findOrCreateUser()
- Updates user name and avatar from Google profile on first Google login
Frontend:
- Add "Continue with Google" button on login page (shown when
NEXT_PUBLIC_GOOGLE_CLIENT_ID is configured)
- Add /auth/callback page to handle Google OAuth redirect
- Add loginWithGoogle to auth store and API client
When repos are present, sub-steps c/d/e/f are now distinct instead of
having two 'c' steps. Each branch (with/without repos) now has its own
complete set of correctly lettered sub-steps.
Prepend the directory of the running multica binary to PATH in the
agent's environment variables. This fixes the issue where isolated
runtimes (e.g. Codex sandbox) cannot find the multica CLI, causing
agent tasks to fail immediately with "command not found: multica".
Closes#451
- Remove duplicate extractOCToolOutput, reuse extractToolOutput from opencode.go
- Rename extractEventText → openclawExtractText to avoid package-level name collisions
- Add clarifying comments for error status stickiness and result event behavior
- Remove redundant extractOCToolOutput tests (already covered by opencode tests)
Add OpenClaw as a fourth supported agent runtime alongside Claude Code,
Codex, and OpenCode. OpenClaw CLI (`openclaw agent -p ... --output-format
stream-json`) is integrated via the same Backend interface pattern.
Changes:
- Add openclawBackend in server/pkg/agent/openclaw.go with NDJSON
event stream parsing (text, thinking, tool_call, error, step, result)
- Register "openclaw" in the agent factory (agent.go)
- Add MULTICA_OPENCLAW_PATH / MULTICA_OPENCLAW_MODEL env var detection
in daemon config
- Include "openclaw" in AGENTS.md config injection alongside codex/opencode
- Add comprehensive unit tests for all event handlers and processEvents
The assignee check in enqueueMentionedAgentTasks silently skipped
explicit @mentions when the target agent was the issue assignee in
a non-terminal status. This broke the review-rejection-retry loop:
when a reviewer rejected a PR and @mentioned the developer agent,
the mention was skipped because the developer was the assignee.
The downstream HasPendingTaskForIssueAndAgent check already prevents
duplicate queued tasks, making the assignee skip redundant. Removing
it ensures explicit @mentions always fire regardless of assignee status.
Closes#431
- Use google/uuid NewV7() for attachment ID and S3 file key instead of
random hex, so the S3 object name matches the attachment record ID
- Add LinkAttachmentsToIssue query to associate orphaned attachments
with a newly created issue
- Pass attachment_ids in CreateIssue request so uploads during issue
creation (before the issue exists) get linked after commit
- Collect and pass attachment IDs in comment-input and reply-input
so comment creation properly links uploaded files
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Keep branch additions (errSilent, exactArgs, examples template blocks)
that were added in the CLI help UX improvement.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add EXAMPLES section to leaf and sub help templates (gh CLI style)
- Add example to attachment download command
- Simplify attachment download description
- Show help output when required args are missing (error first, then help)
- Replace cobra.ExactArgs with custom exactArgs that prints help on failure
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* refactor(cli): overhaul help output to match gh CLI style
- Add gh-style grouped help with CORE/RUNTIME/ADDITIONAL COMMANDS sections
- Use UPPERCASE section headers (USAGE, FLAGS, EXAMPLES, LEARN MORE)
- Format commands as "name: description" with automatic alignment
- Add ENVIRONMENT VARIABLES and EXAMPLES sections to root help
- Apply consistent templates to root, subcommand, and leaf commands
- Update descriptions from "Manage X" to "Work with X" for gh parity
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(execenv): add explicit instruction for agents to always use multica CLI
Agents were using curl/wget to access Multica attachment URLs directly,
which fails due to authentication. Add a prominent "Important" section
to the generated CLAUDE.md template that explicitly prohibits direct
HTTP access and instructs agents to escalate missing CLI functionality
to their workspace owner.
---------
Co-authored-by: Devv <devv@Devvs-Mac-mini.local>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Agents were using curl/wget to access Multica attachment URLs directly,
which fails due to authentication. Add a prominent "Important" section
to the generated CLAUDE.md template that explicitly prohibits direct
HTTP access and instructs agents to escalate missing CLI functionality
to their workspace owner.
- Add gh-style grouped help with CORE/RUNTIME/ADDITIONAL COMMANDS sections
- Use UPPERCASE section headers (USAGE, FLAGS, EXAMPLES, LEARN MORE)
- Format commands as "name: description" with automatic alignment
- Add ENVIRONMENT VARIABLES and EXAMPLES sections to root help
- Apply consistent templates to root, subcommand, and leaf commands
- Update descriptions from "Manage X" to "Work with X" for gh parity
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
When a reply in a thread explicitly mentions only non-agent entities
(members or issues), do not inherit agent mentions from the parent
comment. This prevents false agent triggers when a user is directing
their reply at other people (e.g. "cc @Someone") rather than requesting
work from agents mentioned in the thread root.
Fixes MUL-324
The assignee was unconditionally skipped in the mention path, assuming
on_comment would handle it. But on_comment is suppressed for terminal
statuses (done/cancelled), so an explicit @mention of the assignee had
no effect. Now only skip the assignee dedup when on_comment will
actually fire (non-terminal status).
1. Update CLAUDE.md template to document --limit, --offset, --since
params and guide agents to use pagination when comments are large
2. Add GetJSONWithHeaders to API client; CLI now prints "Showing X of Y
comments" to stderr when paginating
3. Cap --since without --limit at 500 server-side to prevent unbounded
result sets
