createSession captured a TaskStore but submitResponse/retrySession/
rewindSession had no way to receive one — so rehydrated sessions
(loaded from SQLite, no prior createSession call) failed with
"Planning session has no task store and cannot be resumed without
project context" when they needed to rebuild an agent.
- Add optional `store` to Session and capture it on createSession.
- Extend submitResponse, retrySession, rewindSession to accept and
propagate a store argument; ensureSessionAgent falls back to the
session-captured value when the caller doesn't plumb one.
- Update planning.test.ts rehydration paths to pass MOCK_TASK_STORE
where they previously relied on the implicit error to test against.
Fixes 7 failing tests in planning.test.ts (rehydration + retry/rewind
flows). Same pattern applies to mission-interview.ts and the related
session-* test files but is deferred to a separate commit.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
createSession now requires a TaskStore (planning.ts:765); the reconnect
test had a TaskStore in scope but was still passing undefined as the
third arg. Hook it up. Fixes the 1 failing test in this file.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
routes-sandbox-audit.test.ts mocked @fusion/core without
deterministicGuardLocks (causing the workflow-routes import to throw)
and its MockStore lacked getDatabase, which createServer now calls
when wiring ChatStore. Add both. Fixes 3 failing tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous commit collapsed both header spans into a single isMobile-
conditional render, but that broke 3 desktop-header tests that
intentionally assert on both spans being present (the CSS handles
visibility). The real fix for the "Found multiple elements" failures
was the localStorage.clear() in beforeEach; the structural change was
unnecessary. Restore the original dual-render markup.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The header rendered both a "Missions" desktop span AND a mobile span
containing the selected mission title simultaneously, relying on CSS to
hide the inactive variant. jsdom doesn't apply media queries, so
testing-library's getByText found the mission title in both the body
cards and the otherwise-hidden mobile header — failing 49 tests with
"Found multiple elements with the text: <title>". Render only the
active variant based on isMobile.
Also clear localStorage in beforeEach so the SWR mission cache from
prior tests doesn't pre-hydrate into the next render and surface its
own fixtures as duplicates.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The skeleton-loading test failed because earlier tests in the file
populated the SWR cache, so MailboxModal hydrated inbox from
localStorage and skipped the loading skeleton. Reset localStorage in
beforeEach so each test starts with a clean cache.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
MissionManager.cache and MailboxModal.cache tests parsed the raw
localStorage entry as the payload, but writeCache wraps values in
{ savedAt, data }. Pull .data out before asserting on length or
nested fields. Fixes 3 failing tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
swrCache.writeCache wraps values in a { savedAt, data } envelope, but
useChatRooms/useEvals/useInsights/useResearch tests still parsed
localStorage entries as bare arrays. Pull the .data field out before
indexing so the cache assertions hit the actual payload. Fixes 6
failing tests across the four hook test files.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
server.ts createServer now invokes engine.getSelfHealingManager() to
wire the self-healing branch-rebind hook. The headless engine fake in
remote-access-headless.test.ts hadn't been updated. Add a vi.fn() stub
so the mock satisfies the engine surface contract.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
createDatabase(":memory:") fails the absolute-path guard added to the
Database constructor. Use a real tmp dir with inMemory: true, and call
db.init() to materialize project_insight_runs before InsightStore is
constructed. Fixes 5 failing tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When fetchMissions returns a non-array (envelope shape, error path, or
mocked detail object) setMissions(data) leaks the bad shape into render,
where missions.filter() throws "is not a function" and crashes the
entire MissionManager test file. Coerce to [] before storing, matching
the existing defenses on the useState init and cache-hydration paths.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
SettingsModalNodeRouting and settings-mobile both mock ../../api but
omit fetchProjects, which the SettingsModal now calls during mount.
Add a stub returning an empty list. Drops SettingsModalNodeRouting
failures from 14 → 0 and settings-mobile from 9 → 1 (residual failure
is unrelated and tracked separately).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
proxy-routes and routes-agent-skills tests mock @fusion/core but omit
the deterministicGuardLocks Map that register-task-workflow-routes
imports at module-load time. Add the export so the mock satisfies the
shape consumers expect. Fixes 24 failing tests across the two files.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When multiple TaskStore instances watch the same SQLite DB (dashboard +
engine runtime + per-project stores in one dashboard process), every
column move was recorded once per polling instance — inflating
task:moved rows 3x and amplifying failure noise (146k+ moves and 781
task:failed events in 24h against ~165 real done tasks).
Set suppressActivityLogForPollingEmit while checkForChanges fires
re-emit events, and skip the activity-log listeners when it's set. The
originating instance's in-process emit path remains the sole writer.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Completes FN-5153 by delivering the soft-delete verification gate with a new end-to-end test suite covering the full soft-delete lifecycle and a documentation matrix summarizing the gate behavior. Updates the dashboard test file and adds a brief entry to AGENTS.md to surface the new feature.
Fusion-Task-Id: FN-5153
- Add diagnostics.md and FN-4886/FN-5048 entries to docs/README.md index
- Fix test-project.test.ts spy target (updateSettings → init) to match fixture flow
- Remove unused DEFAULT_PROJECT_SETTINGS import and skip empty settings patch
- Add explanatory comments to empty catch blocks in secrets-env-writer
- Refine near-duplicate timestamp guard in triage for cleaner NaN handling
Stabilizes mobile keyboard viewport metrics in the `useMobileKeyboard` hook, with tests covering the hook and `ChatView` integration; a minor dashboard command adjustment is included.
Fusion-Task-Id: FN-5155
