Two iOS Safari-specific fixes for mobile chat:
- ChatView: the main chat keyboard collapsed the instant it opened
because .chat-thread--keyboard-active declared transform/will-change in
CSS, keeping a non-none transform on an ancestor of the focused
composer textarea. iOS blurs a focused input when an ancestor
establishes a transform containing block. Drive the drift transform in
JS only when iOS actually shifts the viewport (offsetTop > 0); the
ancestor stays transform:none on focus so the keyboard stays up.
- QuickChatFAB: the FAB never opened on iPhone because the drag hook
calls setPointerCapture() in pointerdown, which makes WebKit swallow
the synthetic click. Fire the open/close toggle from the drag hook's
pointerup on a tap (a real gesture, so stealth-input focus still raises
the keyboard); keep onClick for mouse/tests with a timer-cleared dedupe.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The mobile chat keyboard would not reliably stay up on iOS. The cause was
the visibilitychange/pageshow 'resync' effect in ChatView: it only ever
ran while the composer was already focused (its activeElement guard), and
on iOS a programmatic focus() fired from setTimeout has no user-gesture
context, so it could not re-raise the keyboard after the blur(). Every
spurious iOS visibilitychange (Control Center, notification banners, app
switches) therefore dismissed the keyboard mid-session and never restored
it.
Remove the blur()+focus() resync. The visualViewport half-state it
targeted is already handled by useMobileKeyboard, which re-snapshots vv
metrics on visibilitychange/pageshow via its settle tail + rAF stability
poll without ever touching textarea focus.
The mobile resync effect did an unconditional textarea blur()+focus() on
every visibilitychange/pageshow across all platforms. On Android,
focus() after blur() cannot re-raise the soft keyboard, so spurious
visibilitychange events (including those fired mid-keyboard-transition)
collapsed the keyboard. Gate the resync to iOS only (its documented
purpose) and only run it when the document is actually becoming visible.
Align the mobile auto-merge integration test viewport mock with screen dimensions.\n\n- Add explicit mobile width and height media-query handling to the viewport mock.\n- Mock window.screen dimensions alongside innerWidth and innerHeight.\n- Restore the original screen object after each test.\n\nFiles changed:\n ...-merge-toggle-blank.mobile-integration.test.tsx | 23 ++++++++++++++++++++++\n 1 file changed, 23 insertions(+)
Fusion-Task-Id: FN-6288
Fusion-Task-Lineage: 3c929d54-b482-41f3-a03e-0a22df0aa9e9
Document how custom providers are configured and surfaced in model pickers.
- Add Custom Providers dashboard guide coverage for adding, editing, deleting, detecting models, and masked API key behavior.
- Explain that saved custom provider models appear in Project Models and workflow model dropdowns.
- Add a stable settings-reference anchor for the customProviders setting.
Files changed:
docs/dashboard-guide.md | 64 ++++++++++++++++++++++++++++++++++++++++++++++
docs/settings-reference.md | 2 +-
2 files changed, 65 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-6286
Fusion-Task-Lineage: 79520376-d448-4048-af43-a532fba68853
Stabilize the self-healing dirty worktree recovery test by avoiding timer and async cleanup races.
- Force real timers for branch conflict reclaim tests.
- Provide explicit integration branch settings and dispatcher behavior for the recovery path.
- Use synchronous fixture setup and cleanup around the recovery patch assertion.
Files changed:
packages/engine/src/__tests__/self-healing.test.ts | 77 +++++++++++++---------
1 file changed, 46 insertions(+), 31 deletions(-)
Fusion-Task-Id: FN-6280
Fusion-Task-Lineage: ef93f14a-1d16-4cca-ac6d-3cfd7249ca21
Use per-invocation Vitest worker roots to keep compound-engineering tests from timing out on stale temp fixtures.
- Allocate a fresh FUSION_TEST_WORKER_ROOT during Vitest global setup and remove it during teardown.
- Preserve per-worker fallback root creation when global setup is not available.
- Clean compound-engineering harness project roots on close and cover the setup invariants with regression tests.
Files changed:
packages/core/src/__test-utils__/vitest-setup.ts | 20 +++++++---
.../core/src/__test-utils__/vitest-teardown.ts | 38 +++++++++++--------
.../src/__tests__/_harness.ts | 7 +++-
.../src/__tests__/setup-invariant.test.ts | 43 ++++++++++++++++++++++
4 files changed, 85 insertions(+), 23 deletions(-)
Fusion-Task-Id: FN-6282
Fusion-Task-Lineage: 8b842dd0-4f51-44de-b2db-8e8bfa97239c
Ensure assigned agents resume task work even when their heartbeat loop is idle.
- add deferred assignment refiring when an agent is assigned work without an active heartbeat run
- track in-process runtime activity so scheduled work is not double-started
- cover idle, active, and concurrent heartbeat assignment paths with scheduler and runtime tests
- document the deferred assignment wake-up behavior and add a changeset
Files changed:
.changeset/fn-6284-deferred-assignment-refire.md | 5 +
docs/agents.md | 2 +
docs/architecture.md | 1 +
.../src/__tests__/heartbeat-scheduler.test.ts | 174 ++++++++++++++++++++-
packages/engine/src/agent-heartbeat.ts | 109 ++++++++++++-
.../runtimes/__tests__/in-process-runtime.test.ts | 24 +++
packages/engine/src/runtimes/in-process-runtime.ts | 3 +
7 files changed, 311 insertions(+), 7 deletions(-)
Fusion-Task-Id: FN-6284
Fusion-Task-Lineage: cf8a9abe-163e-45c6-961a-095891e45ea5
- Drop the unnecessary regex escape in the shell-metacharacter class.
- Move the R17 git-clean post-condition out of finally (no-unsafe-finally): a
dirty tree now fails closed to an inconclusive verdict instead of a
finally-throw that masked the verdict. Test updated to the new behavior.
Adds the app-driving channel (U4 isolated launch + U8 driver, via an injected
seam to keep the engine decoupled from the plugin), a dispatching capability that
routes by assertion shape (test/app/both), the R21 three-outcome mapping
(absent-vs-found by expectation; all infra failures -> inconclusive), and the R20
determinism contract (N-run agreement before an authoritative fail; flaky ->
inconclusive). Channel runs only against the isolated instance (R11).
Replaces the metadata/probe stub with a playwright-core driver (no bundled
browser download; uses a probe-discovered Chrome) exposing navigate/click/type/
observe + idempotent dispose. Browser-unavailable and un-exercisable assertions
return inconclusive (never a false pass/fail); a missing selector returns a
distinct 'absent' negative observation. Driver is a direct export scoped to the
verification run, not registered as a coding-agent tool.
launchIsolatedApp() binds a non-reserved port (reuses the 4040/FUSION_RESERVED_PORTS
guards), serves a freshly-built bundle, runs against a fresh/empty disposable DB
under a run-unique tmpdir with a scrubbed env (no inherited credentials), and tears
down unconditionally on crash/timeout/abort. Process-spawn and bundle-build are
injected seams; 18 tests characterize the R13 isolation contracts. Real end-to-end
launch wires the default spawner behind the seam in a heavier lane.
Widens the stranded-done recovery branch so a reaped task-less validating/done
feature is re-driven to a terminal verdict instead of stalling its slice forever
(closes the reaper->slice-deadlock P0). Adds a Surface Enumeration + adversarial
re-drive reliability suite (git-clean, no duplicate Fix Features, terminal
verdict, no deadlock) gating release. Updates missions docs + CONCEPTS for the
default-to-fail posture, non-mutating verification, and inconclusive verdict.
Adds changeset.
Threads observed-vs-expected failure detail into the generated Fix Feature (R6);
routes inconclusive verdicts to needs-attention with no Fix Feature and a
distinguishable infra event (R21); makes createGeneratedFixFeature idempotent on
(sourceFeatureId, runId) / skip-when-open so recovery/reaper re-drives don't
duplicate features or exhaust the retry budget (R22); persists mission/audit
events for failures, inconclusive, and previously-swallowed triage errors (R16).
The migration-115 bump in U1 left many test files asserting the old version 114;
update them (db-migrate.test.ts now asserts the exported SCHEMA_VERSION constant
to drop it from future sweeps).
The custom provider edit form seeded the API key input with the masked
value (e.g. "abc•••••wxyz") returned by the sanitized GET. Saving or
running "Detect Models" without retyping echoed that mask back — the
probe endpoint rejects it and a persisted mask broke HTTP header
encoding. The field now starts blank with a "leave blank to keep" hint;
an empty field omits apiKey so the stored credential is preserved.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The settings UI displays saved API keys masked with '•' (U+2022). Saving
a provider without retyping the key echoed that mask back, and the PUT
handler persisted it as the real credential. The masked key then flowed
into an Authorization/x-api-key header, throwing "Cannot convert argument
to a ByteString ... value 8226" at request time.
Treat masked values echoed back on update as "unchanged" so the stored
key is preserved, and reject masked values on create/probe. Add tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The /models endpoint restricted results to providers configured in
Fusion's auth.json/models.json stores, but custom providers live in
global settings and register under customProviderRegistryKey(). Their
keys were never added to the allowlist, so their models were filtered
out before reaching the dropdowns. Add the custom provider registry
keys to the configured-providers set.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds a persisted `type` field (behavioral|static) to MissionContractAssertion,
the schema column + migration 115 + version-sweep, and store population with a
conservative static default so existing rows preserve legacy judging. Foundation
for scoping the validator's default-to-fail / verification posture.
Deduplicate self-healing log entries for queued tasks that remain blocked by the same active file-scope overlap.
- Track the last active overlap blocker logged per task and skip duplicate preserved-queued log entries across self-healing passes.
- Clear the memo when blockers change, resolve, or the manager stops so legitimate future recoveries are still logged.
- Cover unchanged blockers, blocker changes, resolved blockers, stop/reset behavior, and FN-5488 stale-blockedBy preservation with regression tests.
Files changed:
packages/engine/src/__tests__/self-healing.test.ts | 123 +++++++++++++++++++++
packages/engine/src/self-healing.ts | 61 ++++++++--
2 files changed, 177 insertions(+), 7 deletions(-)
Fusion-Task-Id: FN-6276
Fusion-Task-Lineage: 169d4402-ec4e-4fa9-8225-2f545c11a3f7
Keep task detail PR and review affordances synchronized with the current project auto-merge setting.
- Thread the live auto-merge value into task detail modals and list split-pane detail views.
- Prefer the live auto-merge setting over the stale fetched modal snapshot while preserving task-level overrides.
- Cover PR and review tab behavior for auto-merge on/off and document the dashboard behavior.
- Evict flaky engine gate entries and add a patch changeset for the published CLI package.
Files changed:
.changeset/fn-6247-automerge-off-modal-stale.md | 5 +
docs/dashboard-guide.md | 1 +
packages/dashboard/app/App.tsx | 3 +-
packages/dashboard/app/components/AppModals.tsx | 2 +
packages/dashboard/app/components/ListView.tsx | 3 +
.../dashboard/app/components/TaskDetailModal.tsx | 4 +-
.../__tests__/TaskDetailModal.create-pr.test.tsx | 171 ++++++++++++++++++++-
packages/engine/vitest.config.ts | 2 -
8 files changed, 182 insertions(+), 9 deletions(-)
Fusion-Task-Id: FN-6247
Fusion-Task-Lineage: 1321c03a-216d-4b15-bf4f-95621d68c9ae
Keep the mobile dashboard pinned while offscreen auto-merge controls are toggled.
- Reset document horizontal scroll during mobile board stabilization and immediately after auto-merge toggles.
- Cover portrait and landscape mobile scroll realignment in the auto-merge integration test.
- Document the real-browser blank-dashboard root cause and add a patch changeset.
Files changed:
.changeset/FN-6243-mobile-auto-merge-blank.md | 5 ++
...bile-auto-merge-toggle-document-scroll-blank.md | 55 ++++++++++++++++++++++
packages/dashboard/app/components/Board.tsx | 47 ++++++++++++++++--
...-merge-toggle-blank.mobile-integration.test.tsx | 49 +++++++++++++++++++
4 files changed, 152 insertions(+), 4 deletions(-)
Fusion-Task-Id: FN-6243
Fusion-Task-Lineage: 91f74de2-667f-4732-a840-47b792288bec
Rescue the already-merged self-healing real-git test while quarantining a separate flaky merger AI suite.
- Call the already-merged recovery path directly in the rescued test and assert specific audit event types instead of exact total counts.
- Remove self-healing-already-merged.real-git.test.ts from the engine default quarantine list and ledger.
- Add merger-ai.test.ts to the engine default quarantine list and quarantine ledger with FN-6238 evidence.
Files changed:
.../self-healing-already-merged.real-git.test.ts | 20 +++++++++++++-------
packages/engine/vitest.config.ts | 2 +-
scripts/lib/test-quarantine.json | 10 +++++-----
3 files changed, 19 insertions(+), 13 deletions(-)
Fusion-Task-Id: FN-6238
Fusion-Task-Lineage: b04451e3-fbd3-4dee-aca2-f542d1036d20
Persist default workflow model lane edits through the Settings modal's primary save path.
- Register a Project Models section save handler so workflow lane overrides flush before closing Settings.
- Surface typed workflow setting rejections without closing the modal or showing a success toast.
- Remove the separate workflow-model save button and update coverage/docs for the unified save behavior.
Files changed:
docs/dashboard-guide.md | 2 +-
docs/settings-reference.md | 17 +++++---
.../dashboard/app/components/SettingsModal.tsx | 11 ++++-
.../components/__tests__/SettingsModal.test.tsx | 50 ++++++++++++++++------
.../settings/sections/ProjectModelsSection.tsx | 42 ++++++++++--------
.../app/components/settings/sections/context.ts | 3 ++
6 files changed, 83 insertions(+), 42 deletions(-)
Fusion-Task-Id: FN-6225
Fusion-Task-Lineage: 98338f85-c742-46c2-8bdf-f676a6f7e32e
Quarantines the flaky active-session pruning coverage without hiding the rest of merger cleanup tests.
- Move the active-session pruning scenario into its own focused test file.
- Keep the broader merger cleanup suite in the reliability project.
- Point the quarantine ledger and engine default exclude at only the isolated flaky test.
Files changed:
.../merger-ai-cleanup-active-session.test.ts | 60 ++++++++++++++++++++++
.../engine/src/__tests__/merger-ai-cleanup.test.ts | 22 ++------
packages/engine/vitest.config.ts | 2 +-
scripts/lib/test-quarantine.json | 4 +-
4 files changed, 66 insertions(+), 22 deletions(-)
Fusion-Task-Id: FN-6207
Fusion-Task-Lineage: b5e806cb-aec8-4893-83ef-d27164250a5c
