- Update lastRefreshAtRef on non-OK /api/sessions responses to prevent
infinite 5-second retry loops when the API is unhealthy
- Reset lastRefreshAtRef when project changes so each project gets its
own immediate first refresh instead of inheriting the prior project's
cooldown timestamp
- Add test for project-switch refresh behavior
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Treat unenriched PRs as a first-class state in the dashboard so SSR can
return fast-path session data without showing misleading PR status, size,
or alert states. Add shimmer placeholders for partial PR data, avoid
classifying unenriched PRs as merge/review/pending work, and ensure the
first SSE snapshot triggers an immediate refresh without failure loops.
- Remove phantom `projects` prop from test setup (was dead code)
- Deduplicate test coverage: consolidate component tests into
OrchestratorSelector.test.tsx, keep only page route tests in
orchestrators.test.tsx
- Fix useRouter mock to capture and assert router.push calls
- Add finally block to handleSpawnNew for resilience (ensures
isSpawning is reset even if router.push fails)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix tmux target to use session ID when runtimeHandle is missing,
preventing references to non-existent sessions
- Extract mapSessionsToOrchestrators helper to eliminate duplicate
orchestrator listing logic between page.tsx and API route
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Export Orchestrator interface from OrchestratorSelector component
and import it in page.tsx to avoid silent divergence risk.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The projects prop was declared and passed but never used, causing
unnecessary data fetching and serialization. Removed the prop from
the component interface, page, and tests.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Guard against NaN timestamps from invalid ISO date strings and handle
future timestamps gracefully by returning "Just now" instead of
negative values.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When ao start is run and existing orchestrator sessions exist for the
project, the CLI now skips spawning a new one. Instead:
- Detects existing orchestrator sessions before spawning
- Opens the dashboard to a new /orchestrators page for session selection
- Users can resume an existing orchestrator or start a new one
Changes:
- packages/cli/src/commands/start.ts: Check for existing orchestrators,
redirect to selection page if found
- packages/web/src/app/api/orchestrators/route.ts: Add GET endpoint to
list orchestrators for a project
- packages/web/src/app/orchestrators/page.tsx: New page for orchestrator
selection
- packages/web/src/components/OrchestratorSelector.tsx: UI component for
selecting or spawning orchestrators
- packages/web/src/components/__tests__/OrchestratorSelector.test.tsx:
Tests for the selector component
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: support multiple concurrent orchestrators with isolated worktrees
- Each orchestrator session gets a numbered ID ({prefix}-orchestrator-N)
and an isolated git worktree, replacing the single shared orchestrator
- reserveNextOrchestratorIdentity atomically reserves the next available
number and detects conflicting project prefix configurations early
- isOrchestratorSession / isOrchestratorSessionName accept optional
allSessionPrefixes for cross-project false-positive prevention
- getProjectPause and resolveGlobalPause track the longest active pause
across all concurrent orchestrators instead of returning the first found
- cleanupWorktreeAndMetadata helper used consistently on all failure paths
including post-launch; system prompt file included in cleanup
- pollBacklog, status.ts, ProjectSidebar, sessions page, global-pause,
project-utils, serialize, and sessions API route all pass
allSessionPrefixes to isOrchestratorSession
- Web sessions/[id]/page.tsx fetches project prefix map and passes it
through prefixByProjectRef to avoid stale closure in fetchProjectSessions
* fix: seed sseAttentionLevels from fresh sessions on full refresh
When scheduleRefresh dispatches a reset action, derive sseAttentionLevels
from the freshly fetched sessions using getAttentionLevel. This clears
phantom entries for removed sessions and seeds correct levels for new
sessions, preventing stale favicon color and attention counts until the
next SSE snapshot.
* fix: merge duplicate @/lib/types imports into single import statement
Combining the separate import type and value import from @/lib/types
into one import to satisfy the no-duplicate-imports lint rule.
Update all usages of project.tracker.plugin and project.scm.plugin
to use optional chaining since the plugin field is now optional
when package or path fields are specified.
Files updated:
- CLI: doctor.ts, status.ts, verify.ts
- Web: issues/route.ts, verify/route.ts, scm-webhooks.ts,
serialize.ts, services.ts
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
tsconfig.server.json sets rootDir:"server" so importing from
../src/lib/mux-protocol.ts violated the boundary and broke the CI
typecheck step. Reverted to local type definitions with a comment
linking to the canonical source.
Also adds tests to bring diff coverage above 80%:
- MuxProvider.test.tsx: 30 tests covering lifecycle, message handling,
and all terminal operations (subscribeTerminal, write, open, close,
resize, reconnect, cleanup)
- api-routes.test.ts: GET /api/sessions/patches coverage (success,
field shape, project filter, error path)
- useSessionEvents.test.ts: mux-path tests (muxSessions snapshot
dispatch, membership-change scheduleRefresh, mux-active SSE bypass,
cleanup on unmount)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After a reconnect MuxProvider re-opens all terminals, but new PTYs spawn
at 80×24 default. DirectTerminal only sent the initial resize on mount.
Add a dedicated effect that fires whenever muxStatus transitions to
"connected": calls fit.fit() to measure the current container then sends
the live cols/rows to the server, keeping the PTY size in sync.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
muxStatus was in the deps array solely for the guard check at the top of
the effect. This caused the entire RAF loop + transitionend + backup timers
to re-run on every mux status transition (e.g. reconnecting→connected),
even when no layout change occurred.
Fix: track muxStatus in a ref (updated on every render) and read the ref
inside the effect, so the guard always sees the current value without
muxStatus being a dependency.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Move buffer send + subscribe inside !subscriptions.has(id) guard so
reconnecting clients don't receive the history replay twice (once on
first open, once after wsRef is reassigned on reconnect)
- Remove `terminals` field from MuxContextValue and useMemo — it was
populated from a write-only ref and never consumed by any component
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- useSessionEvents: mux-active SSE cleanup now also resets
pendingMembershipKeyRef and refreshingRef so the aborted fetch's
.finally() handler cannot reschedule after unmount
- DirectTerminal: add distinct "Disconnected" label and error-coloured
dot for muxStatus "disconnected" (WebSocket constructor failure with
no reconnect), instead of the misleading "Connecting…" shown for
transient states
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The SSE effect had muxSessions (array reference) in its deps, causing
it to re-run on every snapshot and fire cleanup that cleared the
debounce timer the mux effect intentionally preserved.
- Derive muxActive = muxSessions !== undefined and use that in SSE
effect deps — the effect now only re-runs when mux transitions
between present/absent, not on every new array reference
- Add reschedule in scheduleRefresh .finally() abort path: when a
fetch is aborted mid-flight (by a new snapshot) and there is still
a pending membership key, reschedule so the refresh isn't silently
dropped
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- MuxProvider: handle exited and error terminal messages — exited
removes the terminal from openedTerminalsRef (preventing re-open on
reconnect) and writes a red notice into the xterm stream; error is
logged to console
- MuxProvider: remove dead buffersRef/bufferBytesRef — client-side
ring buffer was never read; the server already delivers history on
open
- mux-websocket: use ws.terminate() instead of ws.close() on heartbeat
timeout — an unresponsive peer won't complete the close handshake,
so terminate() immediately destroys the socket and frees resources
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Removing the muxSessions.length === 0 guard so an empty snapshot
triggers the membership-key comparison and scheduleRefresh(), which
fetches the full session list and dispatches a reset. Previously,
[] caused an early return that left removed sessions visible indefinitely.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- MuxProvider: reset isDestroyedRef to false at the start of the
effect so React StrictMode's double-invoke doesn't permanently break
the connection (cleanup sets it true, re-run must reset it)
- DirectTerminal: remove dead status state — it was only set to "error"
in a catch block but never read; displayStatus already derives the
error indicator from the error string directly
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- mux-websocket: wire up the exited ServerMessage — add exitCallbacks
to ManagedTerminal, pass onExit param through subscribe(), and fire
callbacks when PTY exits and re-attach fails so clients get notified
- DirectTerminal: fix displayStatus to show "error" when there is a
local error (e.g. xterm.js load failure) regardless of mux state
- useSessionEvents: SSE effect's mux early-return path now returns a
cleanup function that clears refreshTimerRef and aborts in-flight
requests on unmount, preventing post-unmount dispatch calls
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- DirectTerminal: remove stale muxStatus read from xterm setup effect
(muxStatus was captured at mount, leaving reload button permanently
disabled); reload button now checks muxStatus directly on each render
- TerminalManager: kill PTY and delete map entry when last subscriber
unsubscribes, preventing orphaned node-pty processes when all mux
clients disconnect
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- MuxProvider: assign wsRef.current immediately after construction so
cleanup can close a WebSocket that is still in CONNECTING state;
add isDestroyedRef to prevent the close handler from scheduling
reconnects after the component unmounts
- mux-websocket: client "close" message now only unsubscribes that
client — removed terminalManager.close() which killed the shared PTY
for every other connected client
- useSessionEvents: mux effect cleanup no longer clears the debounce
timer; only aborts in-flight requests, preventing rapid mux snapshots
from starving the membership-change refresh
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace Node.js Buffer.byteLength() with TextEncoder in MuxProvider
(browser-safe UTF-8 byte counting for the ring buffer)
- Remove write-only sessionSubscribedRef from MuxProvider
- Remove dead DirectTerminalLocation, DirectTerminalWsUrlOptions,
buildDirectTerminalWsUrl and their tests (superseded by mux)
- Call closeTerminal(sessionId) on DirectTerminal unmount so PTY
processes are released and openedTerminalsRef stays accurate
- Add cleanup return to mux effect in useSessionEvents so pending
refresh timers and abort controllers are cleared on unmount
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Use native WS ping frames for heartbeat so idle browser clients are not
incorrectly disconnected (browser auto-responds to native ping with pong)
- Pass runtime config to buildMuxWsUrl() so dynamic port/proxy path set via
TERMINAL_WS_PATH env var is actually used (was fetched but never consumed)
- Delay initial WS connect until runtime config fetch resolves, preventing
race condition on first load
- shutdown() now terminates all mux clients and closes the WSS, preventing
orphaned PTY processes on restart
- ws 'error' handler now unsubscribes terminal callbacks alongside session
subscription to prevent leaks in edge cases where 'close' does not follow
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace three separate real-time channels (per-terminal WS, SSE, HTTP poll)
with a single persistent multiplexed WebSocket at /mux.
Architecture:
- Browser ↔ MuxProvider owns one WS connection per tab (/mux)
- Terminal I/O, resize, open/close all flow over mux channels
- Session status patches delivered via a shared SSE relay:
mux server subscribes once to Next.js /api/events (SSE) and
broadcasts to all connected browser clients — no per-client polling
- Manual WS upgrade routing fixes ws library limitation with multiple
WebSocketServer instances on the same HTTP server
Remove:
- terminal-websocket.ts (legacy ttyd-based per-session server, port 14800)
- Per-terminal WebSocket connections from DirectTerminal
- Per-client 5 s HTTP polling for session patches
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implements client-side multiplexed WebSocket context:
- Create MuxProvider component with persistent connection management
- Implement exponential backoff reconnection (1s to 30s)
- Add subscribeTerminal/writeTerminal/resizeTerminal/openTerminal methods
- Manage per-terminal ring buffers (50KB max) for background output
- Track session patches from server
- Create useMux() hook for easy access in components
- Mount MuxProvider in app root layout via Providers wrapper
- Support dynamic runtime config from /api/runtime/terminal
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implements backend multiplexed WebSocket server on /mux endpoint:
- Add mux-protocol.ts with ClientMessage and ServerMessage types
- Create TerminalManager class for managing PTY processes independently
- Implement mux-websocket.ts with attachMuxWebSocket() function
- Wire mux server into existing direct-terminal-ws server
- Support multiple terminals over single persistent WebSocket connection
- Implement 50KB ring buffer for background terminal output
- Add heartbeat and reconnection logic
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: event-driven live tab titles and favicons via SSE
Switch tab titles and favicons from polling to real-time SSE updates:
- Extend useSessionEvents to expose sseAttentionLevels map from SSE
snapshots (server-computed, includes full PR state)
- Refactor DynamicFavicon to use SSE attention levels instead of
recomputing from the full sessions array (which has stale PR data
between refreshes)
- Add useSSESessionActivity hook for session detail page to update
document.title emoji immediately on activity change
- Add live dashboard title showing count of sessions needing attention
- Update PullRequestsPage to use new DynamicFavicon API
Closes#115
* fix: seed initial attention levels to avoid stale favicon/title on first render
Accept initialAttentionLevels parameter in useSessionEvents so callers
can seed the attention map from initialSessions via getAttentionLevel().
This prevents the favicon and dashboard title from briefly showing
"all clear" before the first SSE snapshot arrives.
* fix: add EventSource mock to session page test for SSE hook compatibility
* fix: reset stale activity state when sessionId changes in useSSESessionActivity
Add sessionId to the effect dependency array and reset state to null at
the start of each effect run so callers that reuse the hook with a
different sessionId don't see the previous session's activity.
* fix: reset sseAttentionLevels on initialSessions change to prevent stale data on project switch
The reset action now accepts an optional sseAttentionLevels field. When
initialSessions changes (e.g., project switch via sidebar), the dispatch
passes the current initialAttentionLevels via ref so the favicon and
dashboard title reflect the new project immediately rather than showing
stale attention data until the first SSE snapshot.
* feat: add CI failure detail notifications in lifecycle manager
When CI fails on a PR, the lifecycle manager now fetches individual check
details (names, statuses, URLs) and sends them to the worker session.
This complements the existing static reaction message with actionable
debugging information.
Flow:
- On first transition to ci_failed: static reaction message fires (existing)
- On next poll: detailed CI failure info with check names and URLs dispatched
- Fingerprinting prevents re-sending the same failure set
- New/changed failures trigger fresh detailed notifications
- Tracking metadata cleared when PR is merged/closed or CI passes
Follows the same deduplication pattern as maybeDispatchReviewBacklog().
* fix: send CI details directly to avoid consuming reaction retry budget
The detailed CI failure dispatch now uses sessionManager.send() directly
instead of executeReaction(), so it doesn't increment the ci-failed
reaction tracker. This prevents low retries/escalateAfter settings from
causing premature escalation before the agent receives failure details.
The transition reaction still owns escalation; the detailed dispatch is
purely informational follow-up delivery.
* feat: add merge conflict notifications in lifecycle manager
Adds maybeDispatchMergeConflicts() that detects merge conflicts from
the PR enrichment cache or getMergeability() and notifies the worker
session. Conflicts are dispatched independently of session status since
they can coexist with ci_failed, changes_requested, etc.
- Uses the existing merge-conflicts reaction config
- Dispatches once per conflict occurrence (tracks lastMergeConflictDispatched)
- Clears tracking when conflicts resolve, allowing re-dispatch if they recur
- Sends directly via sessionManager.send() (same pattern as CI details)
* fix: use CICheck type instead of inline type declaration
Replace inline Array<{ name, status, url, conclusion }> with the
existing CICheck type from ./types.js for formatCIFailureMessage and
the checks variable in maybeDispatchCIFailureDetails.
* fix: resolve no-useless-assignment lint error in merge conflict check
Declare hasConflicts without initial value since both branches of the
if/else assign to it before it's read.
* feat: show agent notification state in session page blockers
The blockers section on both SessionDetail (IssuesList) and SessionCard
(alert pills) now shows whether the agent has been notified about each
blocker. Reads lifecycle manager dispatch metadata:
- lastCIFailureDispatchHash for CI failures
- lastMergeConflictDispatched for merge conflicts
- lastPendingReviewDispatchHash for review comments
Displays "agent notified" indicator next to blockers where the lifecycle
manager has already forwarded the issue to the worker session.
* fix: use lifecycle status as fallback for blockers when PR data is stale
The blockers section now uses the lifecycle manager's session status
metadata as a source of truth when PR enrichment data hasn't caught up.
PR enrichment uses a 5-min cache and can timeout or be rate-limited,
causing blockers to show stale/incorrect state.
Changes:
- IssuesList and getAlerts now check metadata["status"] (lifecycle
manager state) alongside PR enrichment data
- CI failing: shown when pr.ciStatus is "failing" OR lifecycle status
is "ci_failed"
- Changes requested: shown when pr.reviewDecision matches OR lifecycle
status is "changes_requested"
- Merge conflicts: shown when pr.mergeability.noConflicts is false OR
lifecycle dispatch metadata indicates conflicts were detected
* fix: fall back to getMergeability when cached hasConflicts is undefined
When PREnrichmentData has hasConflicts as undefined (the field is typed
as boolean | undefined), the previous check treated it as no conflicts.
Now falls through to the getMergeability() call instead.
* test: add coverage for CI/conflict notify action and recovery paths
- Test CI tracking clears when CI recovers to passing
- Test notify action for CI failure details (human notification path)
- Test notify action for merge conflicts (human notification path)
These cover the previously uncovered notify action branches and the
CI recovery cleanup path in the lifecycle manager.
* fix: resolve typecheck error in CI recovery test
writeMetadata requires SessionMetadata type which doesn't include
custom keys like lastCIFailureFingerprint. Rewrote the test to use
setupCheck and let the lifecycle manager set tracking metadata
naturally through the CI failure flow, then verify cleanup on recovery.
* fix: don't use dispatch metadata for conflict detection in UI
lastMergeConflictDispatched lingers after conflicts resolve until the
lifecycle manager's next poll clears it. Using it as a conflict signal
caused stale "merge conflict" alerts. Now only pr.mergeability.noConflicts
drives conflict detection; the metadata is only used for the "agent
notified" badge.
* fix: use Promise.allSettled for dispatch functions to avoid orphaned rejections
Promise.all rejects immediately on first failure, leaving in-flight
promises unmonitored. Promise.allSettled waits for all to complete.
Ashish Huddar
harshitsinghbhandari
Harsh Batheja
Gaurav Bhola
Tanuu