Extracted the duplicated file-watch-with-polling-fallback logic from TaskStore and AgentStore into a shared controller.
- Added packages/core/src/fs-watch-poll-controller.ts implementing a reusable fs.watch + polling-fallback controller
- Refactored packages/core/src/store.ts (TaskStore) to use the shared controller instead of its own watch/poll implementation
- Refactored packages/core/src/agent-store.ts (AgentStore) to use the shared controller instead of its own watch/poll implementation
- Added packages/core/src/__tests__/fs-watch-poll-controller.test.ts covering the new controller's behavior
- Updated docs/architecture.md to document the shared controller
Files changed:
docs/architecture.md | 1 +
.../src/__tests__/fs-watch-poll-controller.test.ts | 187 +++++++++++++++++++++
packages/core/src/agent-store.ts | 66 +++-----
packages/core/src/fs-watch-poll-controller.ts | 123 ++++++++++++++
packages/core/src/store.ts | 66 +++-----
5 files changed, 364 insertions(+), 79 deletions(-)
Fusion-Task-Id: FN-7726
Fusion-Task-Lineage: 69be6dc3-5414-44f2-a3f1-3eb72c2d7391
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Fixes a task_agent_mutation policy example drift: the read-only GitLab browse
tools were incorrectly listed as mutation examples even though they were
never part of ACTION_GATE_TASK_AGENT_MANAGEMENT_TOOLS and are already
classified read-only.
- Remove fn_task_browse_gitlab_project_issues, fn_task_browse_gitlab_group_issues,
and fn_task_browse_gitlab_merge_requests from AGENT_PERMISSION_POLICY_CATEGORY_TOOL_EXAMPLES.task_agent_mutation
in packages/core/src/types.ts, keeping the mutating fn_task_import_gitlab_* variants
- Add an FNXC:ToolGovernance comment documenting the invariant and rationale
- Add regression coverage asserting the browse tools are excluded from
task_agent_mutation examples and are pinned as READONLY_FN_TOOLS
Files changed:
.../src/__tests__/agent-permission-policy.test.ts | 19 +++++++++++++++++++
packages/core/src/types.ts | 6 ++----
2 files changed, 21 insertions(+), 4 deletions(-)
Fusion-Task-Id: FN-7733
Fusion-Task-Lineage: 3a54faa1-89dd-48bc-978a-4a53f06706be
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Board mutations (fn_task_update, CEO override, direct SQL) issued from a pi-extension tool session could silently write into a task's throwaway, never-synced worktree-local .fusion/fusion.db instead of the true project root when git CLI resolution failed (missing git binary, Docker "dubious ownership" refusal, or a non-default settings.worktreesDir). This fixes root-cause resolution and adds regression coverage plus a docs writeup.
- getProjectRootFromGitLinkedWorktree now resolves a linked worktree's project root from git's own on-disk .git/commondir metadata via pure filesystem reads before falling back to the git rev-parse CLI, so writes no longer fall through to a local hydrated copy on git-invocation failure.
- Added getMainRepoRootFromGitFile and resolveCommonGitDirFromWorktreeGitFile helpers with FNXC:Storage comments documenting the FN-7730 root cause and fix rationale.
- Added packages/core/src/__tests__/pi-extensions-write-path-durability.test.ts regression coverage for the write-path durability invariant.
- Extended packages/core/src/__tests__/pi-extensions.test.ts with additional resolution-path assertions.
- Documented the failure mode and fix in docs/storage.md ("Silent board-mutation write loss (FN-7730)").
- Added a patch changeset for @runfusion/fusion describing the user-facing fix.
Files changed:
.changeset/fn-7730-worktree-project-root-resolution.md | 7 ++
docs/storage.md | 54 ++++++++++
packages/core/src/__tests__/pi-extensions-write-path-durability.test.ts | 98 ++++++++++++++++++
packages/core/src/__tests__/pi-extensions.test.ts | 87 +++++++++++++++-
packages/core/src/pi-extensions.ts | 114 +++++++++++++++++++++
5 files changed, 359 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-7730
Fusion-Task-Lineage: 00753a2d-a934-42cf-8fde-0f9b8ad98142
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Introduces a dedicated review_gate_bypass permission-policy category so operators can govern who may bypass a failed pre-merge review gate independently of ordinary task-mutation permissions.
- Add review_gate_bypass as a new sensitive-action category in packages/core/src/types.ts, distinct from task_agent_mutation, with fn_task_bypass_review as its sole example tool
- Default review_gate_bypass to require-approval even under the unrestricted preset (stricter than the preset's uniform disposition) in packages/core/src/agent-permission-policy.ts, while approval-required/locked-down already cover it uniformly
- Classify fn_task_bypass_review into the new category via a shared REVIEW_GATE_BYPASS_FN_TOOLS set in packages/engine/src/gating-classifications.ts, consumed identically by both evaluateAgentActionGate and the permanent-agent gate to prevent path drift
- Render the new category as its own row in the dashboard's project-default and per-agent AgentPermissionPolicyEditor, surfaced in AgentDetailView
- Update docs/settings-reference.md and add unit tests across core/engine/dashboard covering the new category, its stricter default, and gate-classification alignment
- Add changeset (@runfusion/fusion: minor) documenting the new operator-facing permission category
Files changed:
.changeset/fn-7728-review-gate-bypass-rbac.md | 7 +++
docs/settings-reference.md | 8 +--
.../src/__tests__/agent-permission-policy.test.ts | 54 ++++++++++++++++++-
packages/core/src/agent-permission-policy.ts | 12 ++++-
packages/core/src/types.ts | 8 +++
.../dashboard/app/components/AgentDetailView.tsx | 2 +
.../app/components/AgentPermissionPolicyEditor.tsx | 8 +++
.../__tests__/AgentPermissionPolicyEditor.test.tsx | 5 ++
.../engine/src/__tests__/agent-action-gate.test.ts | 45 ++++++++++++++++
.../src/__tests__/gating-classifications.test.ts | 63 ++++++++++++++++++++++
.../src/__tests__/permanent-agent-gating.test.ts | 41 ++++++++++++++
packages/engine/src/agent-action-gate.ts | 7 +++
packages/engine/src/gating-classifications.ts | 8 ++-
packages/engine/src/permanent-agent-gating.ts | 6 +++
14 files changed, 266 insertions(+), 8 deletions(-)
Fusion-Task-Id: FN-7728
Fusion-Task-Lineage: 100c8563-2897-4d53-9546-5c2faa6ab7d8
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Add a policy-gated review-lane bypass primitive so operators can unstick cards stranded by a failed pre-merge review step (e.g. the no-feedback review-engine defect), without exposing it to agent-driven lanes.
- Add `store.bypassFailedPreMergeReviewStep(id, { reason, actor })` in @fusion/core plus `getLatestFailedPreMergeReviewStep` in task-merge.ts, and new `bypassedBy`/`bypassedAt`/`bypassReason`/`bypassedFromStatus`/`bypassedFromVerdict` fields on `WorkflowStepResult`
- Add operator-only `fn_task_bypass_review` CLI/pi-extension tool; explicitly withheld from executor/reviewer/triage agent tool lists
- Add `POST /tasks/:id/bypass-review` dashboard API route and wire it through `register-task-workflow-routes.ts` and legacy API compatibility layer
- Add dashboard UI affordance (context menu action + task detail modal + right-dock controller wiring) to trigger the bypass with a reason
- Add i18n strings for the bypass action/labels across en/es/fr/ko/zh-CN/zh-TW locales
- Update `gating-classifications.ts` to recognize the bypassed state
- Add unit tests: `store-bypass-review.test.ts`, `task-merge-bypass.test.ts`, extension test coverage, and `useTasks` hook test coverage
- Update docs (`docs/workflow-steps.md`, `docs/dashboard-guide.md`, AGENTS.md, fusion skill references) to describe the new bypass tool/route
- Add changeset `.changeset/fn-7720-review-lane-bypass-primitive.md` (minor)
Files changed:
$(git diff --cached --stat)
Fusion-Task-Id: FN-7720
Fusion-Task-Lineage: 590b020a-ae02-4b51-8189-df8f54bf3044
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Adds opt-in cross-process change detection to AgentStore so the engine reacts to CLI-driven agent stop/start mutations promptly instead of waiting for the periodic audit sweep.
- AgentStore gains fs.watch-based (with poll fallback) cross-process notification, modeled on TaskStore's existing mechanism
- Re-emits existing agent:updated/agent:stateChanged events in the engine process when another process (the fn CLI) mutates an agent row
- HeartbeatTriggerScheduler listeners now fire immediately instead of waiting up to 60s for the auditTimerRegistrations sweep; the sweep remains as durable backstop
- in-process-runtime.ts wires up the new notification bus
- Adds unit tests for agent-store cross-process notifications and heartbeat-scheduler reaction behavior
- Updates docs/agents.md and docs/architecture.md
- Adds changeset (patch) for @runfusion/fusion
Files changed:
.changeset/fn-7723-cross-process-agent-notify.md | 7 +
docs/agents.md | 1 +
docs/architecture.md | 1 +
packages/core/src/__tests__/agent-store.test.ts | 177 +++++++++++++++++
packages/core/src/agent-store.ts | 210 ++++++++++++++++++++-
.../src/__tests__/heartbeat-scheduler.test.ts | 162 ++++++++++++++++
packages/engine/src/runtimes/in-process-runtime.ts | 30 +++
7 files changed, 587 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-7723
Fusion-Task-Lineage: d3a7fa05-b40d-4388-8e98-140f9d8861c9
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Bounds durable-agent heartbeat worktree acquisition to a fixed retry count instead of requeuing to todo indefinitely across heartbeat cycles.
- Add MAX_HEARTBEAT_WORKTREE_ACQUISITION_RETRIES (3) in agent-heartbeat.ts, reusing Task.recoveryRetryCount as a cross-heartbeat counter (no schema migration)
- On cap exhaustion, terminally mark the task status:"failed" with an explanatory error, log the entry, and reopen to todo with preserveStatus so the failed status isn't wiped by reopen-to-todo semantics
- Add onTaskAcquisitionExhausted callback wired in in-process-runtime.ts to CentralCore.recordTaskCompletion(taskId, false) so exhausted acquisitions count toward totalTasksFailed
- Add regression tests in agent-heartbeat-worktree.test.ts and in-process-runtime.test.ts covering the retry cap and completion recording
- Add changeset (patch) and a docs/solutions/logic-errors writeup documenting the investigation and other worktree-collision sub-gaps found not to reproduce on HEAD
Files changed:
.changeset/fn-7721-worktree-heartbeat-retry-cap.md | 7 ++
docs/solutions/logic-errors/heartbeat-worktree-acquisition-unbounded-requeue.md | 84 ++++++++++++++++++++++
packages/engine/src/__tests__/agent-heartbeat-worktree.test.ts | 58 +++++++++++++++
packages/engine/src/__tests__/in-process-runtime.test.ts | 11 +++
packages/engine/src/agent-heartbeat.ts | 72 ++++++++++++++++++-
packages/engine/src/runtimes/in-process-runtime.ts | 12 ++++
6 files changed, 242 insertions(+), 2 deletions(-)
Fusion-Task-Id: FN-7721
Fusion-Task-Lineage: caad671c-f360-4c1c-8aaa-5b48fca5a55b
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Ensures stopping and restarting an agent durably clears its heartbeat timer instead of relying on the later FN-7645 watchdog repair.
- HeartbeatTriggerScheduler.auditTimerRegistrations now unregisters lingering timers for non-eligible (stopped/paused/disabled) agents
- syncTimerForAgent force-re-arms a stale present timer on a start transition so no orphaned timer entry lingers
- Added 308 lines of new heartbeat-scheduler regression tests covering the stop/start zombie-timer scenarios
- Added changeset (patch) documenting the fix
- Updated docs/agents.md and docs/architecture.md to describe the new invariant
Files changed:
.changeset/fn-7718-zombie-timer-invalidate.md | 7 +
docs/agents.md | 2 +
docs/architecture.md | 1 +
.../src/__tests__/heartbeat-scheduler.test.ts | 308 +++++++++++++++++++++
packages/engine/src/agent-heartbeat.ts | 49 +++-
5 files changed, 364 insertions(+), 3 deletions(-)
Fusion-Task-Id: FN-7718
Fusion-Task-Lineage: fc834ccd-495e-4294-805d-325b4cb536a2
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Triage planning-model retries no longer dead-end on "no fallback configured" when a provider primary-model call fails (e.g. a 404 wrapped in a 429 for nvidia/moonshotai/kimi-k2.6).
- Add resolveImplicitPlanningFallbackModel() to agent-session-helpers.ts: derives a fallback from the resolved project/global default (execution) model when neither planningFallback*/global fallback* is set, guarding against self-swap and skipping in test mode.
- Wire the implicit fallback into TriageProcessor.specifyTask() in triage.ts so a retryable primary planner-model failure swaps once via the derived fallback instead of failing triage outright.
- Add unit test coverage in agent-session-helpers.test.ts and triage.test.ts for the new implicit-fallback resolution and its triage integration.
- Document the new implicit-fallback behavior in docs/settings-reference.md.
- Add a patch changeset for @runfusion/fusion describing the fix.
Files changed:
.changeset/fn-7719-triage-planning-implicit-fallback.md | 7 +
docs/settings-reference.md | 2 +
packages/engine/src/__tests__/agent-session-helpers.test.ts | 75 +++++++
packages/engine/src/__tests__/triage.test.ts | 237 +++++++++++++++++++++
packages/engine/src/agent-session-helpers.ts | 41 ++++
packages/engine/src/triage.ts | 31 ++-
6 files changed, 389 insertions(+), 4 deletions(-)
Fusion-Task-Id: FN-7719
Fusion-Task-Lineage: 69e797e1-5bac-47f3-8dce-505b9d64d83c
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Grok CLI provider readiness now mirrors the Cursor CLI provider: it is derived from the `grok` binary being available rather than requiring a Fusion-visible GROK_API_KEY or ~/.grok/user-settings.json, since the CLI manages its own auth.
- probeGrokBinary now derives `authenticated` from binary availability (readiness) instead of API-key/user-settings presence; key detection surfaces as a non-blocking `apiKeyDetected` hint
- /auth/status treats the grok-cli provider as authenticated when enabled + binary available
- GrokCliProviderCard drops the blocking "Set GROK_API_KEY" state
- Direct xAI streaming path is unchanged and still uses $GROK_API_KEY when present (FN-7711/FN-7714)
- Added changeset for @runfusion/fusion (patch)
Files changed:
$(cat /tmp/diffstat_fn7716.txt)
Fusion-Task-Id: FN-7716
Fusion-Task-Lineage: ac0efc79-2510-465e-9cd2-4938c08989c9
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Archiving a task from triage/planning/todo (not just in-progress) previously left leaked active-session-registry entries, so a successor task could hit ActiveSessionPathHeldByForeignTaskError and get blocked from Plan Review.
- Add an explicit `to === "archived"` branch in the task-move handler that awaits abort of in-flight task work and sweeps any leftover activeSessionRegistry paths for the task, checked before the narrower `from === "in-progress"` branch so direct in-progress→archived transitions are covered too.
- Deliberately exclude `to === "done"` / `to === "in-review"` from this sweep since those columns legitimately hold ai-merge / workspace-repo-land merge leases that must survive the transition.
- Add regression test coverage for archive releasing active sessions across originating columns.
- Add changeset and architecture doc note.
Files changed:
.../fn-7717-archive-active-session-release.md | 7 +
docs/architecture.md | 1 +
...xecutor-archive-releases-active-session.test.ts | 167 +++++++++++++++++++++
packages/engine/src/executor.ts | 35 +++++
4 files changed, 210 insertions(+)
Fusion-Task-Id: FN-7717
Fusion-Task-Lineage: 7cff6821-7bb3-4b75-b502-a26467ca7f51
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Fall back to the Grok CLI's user-settings file for the API key so pi's $GROK_API_KEY provider reference resolves even when the env var isn't exported.
- Add hydrateGrokApiKeyFromUserSettings() in grok-provider.ts, called from registerBuiltInGrokProvider(), which hydrates process.env.GROK_API_KEY from ~/.grok/user-settings.json { apiKey } only when the env var is unset/empty
- Env var always wins; a missing (ENOENT), malformed, or empty-apiKey settings file is fail-soft (no throw, no env mutation), mirroring the grok-runtime probe's fallback behavior
- Add regression tests covering env-precedence, fallback hydration, and fail-soft error paths (grok-provider-user-settings.test.ts)
- Document the fallback in docs/settings-reference.md
- Add a patch changeset for @runfusion/fusion
Files changed:
.changeset/fn-7714-grok-user-settings-apikey.md | 7 +
docs/settings-reference.md | 2 +-
.../__tests__/grok-provider-user-settings.test.ts | 156 +++++++++++++++++++++
packages/core/src/grok-provider.ts | 47 +++++++
4 files changed, 211 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-7714
Fusion-Task-Lineage: 5450b480-3a32-4331-9494-867b84605464
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Agents that must edit files beyond a task's declared ## File Scope had no
way to keep the scope in sync, so those edits were stranded at merge (the
squash merge is scoped to ## File Scope, and cross-task overlap blocking +
the merge file-scope invariant both read it).
New executor tool fn_task_file_scope_add validates repo-relative
paths/globs with isValidFileScopeEntry, de-dupes against existing scope,
appends them to the ## File Scope section of PROMPT.md, and persists via
store.updateTask({ prompt }) (same validation + task.json/PROMPT.md sync as
fn_task_prompt_write). Registered in the main coding-agent tool list; the
base executor prompt now instructs the agent to call it when editing beyond
the declared scope. Merge-time peer-claim refusal is unchanged and remains
the cross-task backstop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
FN-7690 changed resolveApiType() to return the registered pi-ai key
"anthropic-messages" for anthropic-compatible providers (the bare
"anthropic" key is never registered and throws at stream time), but left
behind a stale JSDoc and a stale test expectation:
- custom-provider-registry.ts: update the FN-7689 buildCustomProviderModels
comment that still described the anthropic/anthropic-messages drift as
unresolved.
- provider-registration.test.ts: assert config.api === "anthropic-messages"
(was still asserting the pre-fix "anthropic").
Also de-slow a retry-exhaustion test: the describe uses fake timers with
shouldAdvanceTime, so awaiting a 3-retry backoff (1s+5s+15s) burned ~21s of
real wall time. Drive the backoff with advanceTimersByTimeAsync instead
(Standing Rule: prefer fake timers over real time waits).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Prevents short-lived CLI processes from being held open by background SQLite integrity checks.
- unref the sqlite3 child process (and its stdio) spawned by integrityCheckSqliteFileAsync via the shared unrefQmdChildProcess helper, immediately after spawn
- unref the 60s scheduling timer in scheduleBackgroundIntegrityCheck so a short-lived caller isn't pinned waiting for a background check it never asked to block on
- add regression test coverage (db-integrity-check-unref.test.ts) plus a CLI fixture (db-integrity-check-fixture.mjs) that exercises the fix in a real short-lived process
- add changeset documenting the fix and the audit of other spawn sites across @fusion/core/@fusion/engine/@fusion/dashboard/cli confirming they are safe
Files changed:
.changeset/fn-7709-db-integrity-check-unref.md | 7 ++
.../src/__tests__/db-integrity-check-unref.test.ts | 135 +++++++++++++++++++++
.../fixtures/db-integrity-check-fixture.mjs | 28 +++++
packages/core/src/db.ts | 26 ++++
4 files changed, 196 insertions(+)
Fusion-Task-Id: FN-7709
Fusion-Task-Lineage: 6594aca4-0268-4bba-9a7f-af96d695f1e9
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
The OAuth expiry monitor and validity logger iterated the un-aliased
getOAuthProviders() id `anthropic` and evaluated get("anthropic"), which
can resolve to a stale legacy/supplemental row (e.g. ~/.pi/agent/auth.json)
even when the fresh, actually-used token lives under `anthropic-subscription`.
That fired a false "Anthropic OAuth expired" notification while the real
subscription token had refreshed successfully.
Both surfaces now resolve the freshest of the two aliased ids via a shared
resolveEffectiveOAuthCredential helper (mirroring the refresh scheduler's
getRefreshCandidateIds alias handling), so a live subscription token
suppresses the false alert. Notification throttle/cadence unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Fixes searchWithQmd's inline promisify(execFile) copy that could hold a caller open by reusing the already-hardened, synchronously-unref'd executor established for the background refresh path.
- searchWithQmd now calls getDefaultExecFileAsync() instead of building its own promisify(execFile) executor inline
- Removes the second un-unref'd execFile executor that could keep a short-lived caller (e.g. one-shot CLI memory search) open up to the awaited timeout
- Adds regression test fixture and test coverage (qmd-search-fixture.mjs, qmd-search-unref.test.ts) asserting the shared executor is used
- Adds changeset (patch) for @runfusion/fusion
Files changed:
.changeset/fn-7707-qmd-search-unref.md | 7 +
packages/core/src/__tests__/fixtures/qmd-search-fixture.mjs | 30 ++++
packages/core/src/__tests__/qmd-search-unref.test.ts | 166 +++++++++++++++++++++
packages/core/src/memory-backend.ts | 16 +-
4 files changed, 216 insertions(+), 3 deletions(-)
Fusion-Task-Id: FN-7707
Fusion-Task-Lineage: 3f9f94a7-5613-4b9a-a3ba-8e9bcdd6b687
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Replaces promisify(execFile) with a hand-rolled spawn()-based qmd executor that unrefs the child process and its stdio, so a fire-and-forget scheduleQmd* memory-index refresh never keeps a short-lived caller process (e.g. CLI) alive; long-lived callers like the dashboard server still see refresh resolve/reject normally.
- memory-backend.ts: replace promisify(execFile) qmd exec path with spawn()-based executor that unrefs child + stdio
- Add qmd-refresh-unref.test.ts covering unref behavior with a qmd-refresh-fixture.mjs test fixture
- Add changeset fn-7706-qmd-unref.md (patch, fix category)
Files changed:
.changeset/fn-7706-qmd-unref.md | 7 ++
.../src/__tests__/fixtures/qmd-refresh-fixture.mjs | 28 +++++
.../core/src/__tests__/qmd-refresh-unref.test.ts | 136 +++++++++++++++++++++
packages/core/src/memory-backend.ts | 123 ++++++++++++++++++-
4 files changed, 291 insertions(+), 3 deletions(-)
Fusion-Task-Id: FN-7706
Fusion-Task-Lineage: 713c23c2-d7da-42c4-b066-31883ba78321
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Fix CLI process exit so `fn agent stop`/`fn agent start` no longer hang up to 60s and eventually time out on repeated retries against the same agent.
- Root cause: `resolveProject()` cached an unclosed `TaskStore`, and `createAgentStore()` never closed the `AgentStore` it opened, leaving SQLite handles alive after the command's real work was done.
- Add `resolveProjectPathOnly`/`closeProjectStore` helpers in `project-context.ts` so path-only callers never leak a `TaskStore`.
- Explicitly close `AgentStore` on every exit/return path in `agent.ts`, since `process.exit()` skips pending `finally` blocks.
- Add a bounded fast-fail timeout around the state-store write (default 10s, overridable via `FUSION_AGENT_CMD_TIMEOUT_MS`) so a genuinely stuck operation fails fast with a clear error and non-zero exit instead of hanging.
- Add regression tests covering process-exit/store-closing behavior and update CLI reference docs.
- Add changeset for the patch release.
Files changed:
.changeset/fn-7704-agent-cmd-hang-fix.md | 7 +
docs/cli-reference.md | 3 +
.../commands/__tests__/agent-process-exit.test.ts | 114 +++++++++++
packages/cli/src/commands/__tests__/agent.test.ts | 111 +++++++++-
packages/cli/src/commands/agent.ts | 223 ++++++++++++++++-----
packages/cli/src/project-context.ts | 44 ++++
6 files changed, 444 insertions(+), 58 deletions(-)
Fusion-Task-Id: FN-7704
Fusion-Task-Lineage: 4679d1a0-3ab8-48ce-86b7-5919bba805fb
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Fixes the search icon overlapping placeholder/typed text in the Files — Project search input under the compact spacing theme.
- Anchor .file-browser-search-input padding-left to the icon's own --space-sm offset + 16px icon width + a real gap, instead of the unrelated calc(--space-lg + --space-md) formula that collided exactly with the icon's occupied width under compact spacing
- Add FNXC:FileBrowser comment documenting the collision math and why the padding is now theme-invariant
- Add a regression test asserting padding-left exceeds icon offset + width against the compact spacing scale
- Add a patch changeset for @runfusion/fusion
Files changed:
.changeset/fn-7703-search-icon-overlap.md | 7 +++
packages/dashboard/app/components/FileBrowser.css | 11 ++++-
.../app/components/__tests__/FileBrowser.test.tsx | 52 ++++++++++++++++++++++
3 files changed, 69 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-7703
Fusion-Task-Lineage: d7de7f82-f0e7-4086-a8ef-2fed2b4704ec
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Thread the machine-local cursorCliBinaryPath operator override into /api/models Cursor CLI discovery so the model picker spawns the same cursor-agent binary already validated by sign-in/status/probe.
- register-model-routes.ts reads globalSettings.cursorCliBinaryPath, trims and normalizes blank to undefined (preserving PATH auto-detection)
- passes the normalized binaryPath through to getCursorPickerModels({ binaryPath }) for model discovery
- adds regression tests covering override-set and override-blank/unset behavior in register-model-routes-cursor-cli.test.ts
- adds changeset fn-7699-cursor-cli-binary-path-model-picker.md (patch, fix) documenting the follow-up to FN-7696
Files changed:
$(cat /tmp/diffstat.txt)
Fusion-Task-Id: FN-7699
Fusion-Task-Lineage: 9895af8e-447d-425b-af58-1af4748c013c
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Adds an end-to-end integration test proving the artifact pipeline works against a real TaskStore, not just a mocked one.
- New test file exercises createArtifactRegisterTool/createArtifactListTool/createArtifactViewTool bound to a real TaskStore (inMemoryDb, real filesystem writes) instead of a mocked store
- Pins the register -> list -> view invariant for a real base64 PNG image artifact, verifying disk persistence, SQLite row fields (type, mimeType, sizeBytes, uri, taskId), and the list/view text surfaces
- Pins the invalid-base64-payload rejection path (non-image bytes for an image-typed artifact) to confirm no artifact row is persisted
- Pins the empty-state list text for a task with no registered artifacts
Files changed:
packages/engine/src/__tests__/agent-artifact-tools-real-store-integration.test.ts | 131 +++++++++++++++++++++
1 file changed, 131 insertions(+)
Fusion-Task-Id: FN-7693
Fusion-Task-Lineage: fd3493aa-6fb2-4e18-a735-c4a9d87c9c6c
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
The 'publishes failed terminal workflow step activity' test awaited executeAll()
directly while the executor retried a failing step 3x with sleep() delays. Under
useFakeTimers({ shouldAdvanceTime: true }) those sleeps consumed REAL wall-clock
time (~22.6s locally, ballooning under CI load and busting the shard-2 watchdog).
Fast-forward the retry sleeps via vi.advanceTimersByTimeAsync like sibling retry
tests; the loop now completes in milliseconds.
## Summary
Follow-up to #1947. The full-suite on `main` is still red on 3 surfaces
introduced by post-#1947 commits. This PR fixes the two real test
failures and the i18n parity gap.
## Fixes
- **i18n key parity (FN-7658)** —
`settings.scheduling.autoArchiveDuplicateTasks` + `...Help` were added
to `en` but not the 5 non-en catalogs, breaking the i18n parity gate
(`parity.test.ts`, `i18n-gate-coverage.test.ts`). Added the 2 keys
(empty-string per the untranslated-entry convention) to `zh-CN`,
`zh-TW`, `fr`, `es`, `ko` in `packages/i18n/locales` (the single source
of truth; `dashboard/app/locales` is gitignored and synced in CI).
- **chat.test.ts (FN-7675)** — `chat.ts` now imports
`FUSION_RUNTIME_SELF_AWARENESS` from `@fusion/core` (CHAT_SYSTEM_PROMPT
embeds it); the hand-written core mock didn't stub it, so the module
failed to load. Added a stub (importOriginal intentionally avoided to
preserve the fs-cascade block).
- **verification-followup-dedup.test.ts (FN-7658)** — the "remains
additive with FN-4892 same-agent duplicate intake" test asserts the
ARCHIVE path, but FN-7658 made same-agent auto-archiving opt-in
(`autoArchiveDuplicateTasksEnabled` defaults false → flag-in-place in
triage). The test now opts into the legacy archive behavior it asserts.
## Note on shard-2 engine[2/2] timeout
The full-suite shard 2 times out on `@fusion/engine [2/2]` (watchdog
900s). Locally `[2/2]` runs in ~96s and is green (the lone
`provider-registration.test.ts` failure is local-only `pi-ai@0.79.9`
staleness — the lockfile pins `0.80.3` which exports `/compat`, so CI
resolves it). The `verification-followup-dedup` failure above is the
only real `[2/2]` defect; this PR fixes it. If the CI timeout persists
it's aggregate real-git load, which I'll address separately (not a code
bug).
## Verification
- i18n `parity` + `i18n-gate-coverage`: 7/7 ✅
- `chat.test.ts`: 14/14 ✅
- `verification-followup-dedup`: 5/5 ✅
- engine `--shard=2/2` (excluding the local-staleness file): 363 files /
4483 tests ✅ in ~96s
No production behavior change; no changeset needed (i18n catalog +
test-only).
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added scheduling settings for automatic duplicate-task archiving,
including label and help text entries (currently placeholders) across
Spanish, French, Korean, Simplified Chinese, and Traditional Chinese.
* **Bug Fixes**
* Updated “awaiting confirmation” merger messaging to better reflect
when auto-merge proceeds automatically.
* **Tests**
* Updated reliability interaction tests to explicitly opt into legacy
duplicate-task archiving behavior.
* Adjusted chat-related tests by extending the runtime mock to satisfy a
new core import requirement.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Follow-up hardening on the RuntimeFallbackBadge viewport-gating work:
- AgentsView.tsx: registerAgentCardRef now returns a cached, stable callback per
key (agentCardRefCallbacksRef) instead of a fresh closure each render. A fresh
closure reads as unmount+remount to React; in environments without
IntersectionObserver the mount path calls setVisibleAgentCardKeys -> re-render
-> another fresh closure -> an infinite re-render loop (including jsdom). The
cached entry is evicted on true unmount (el === null) so the Map cannot grow
unbounded across created/deleted agents.
- ActiveAgentsPanel.tsx: document the viewport-gated badge polling with an FNXC
comment (behavior unchanged).
- useRuntimeFallbackStatus.ts: guard __resetRuntimeFallbackToastDedupeStoreForTests
to a no-op outside the test build (import.meta.env.MODE !== "test") so the
test-only dedupe reset can never affect production code paths.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Real root cause of the blank mobile terminal (the FN-7692 remeasure guard did not
fix it and is reverted here). styles.css has a mobile-only reset
`@media (max-width: 768px) { * { max-width: 100% } }` to prevent horizontal
overflow. That universal selector also matches xterm's hidden character-measurement
subtree (`.xterm-helpers` / `.xterm-char-measure-element`). That subtree's containing
block (`.xterm-helpers`) is a 0x0 absolutely-positioned box, so `max-width: 100%`
resolves to `max-width: 0` and hard-caps xterm's character-cell measurement at 0.
FitAddon.fit() then proposes 0 columns/rows and `.xterm-screen` (plus the WebGL
canvas) collapses to 0x0 — the prompt streams in and is written into xterm's row DOM
but paints into a zero-size box, so the terminal is blank. Mobile-only, which is why
desktop always rendered fine.
Reproduced live via mobile emulation: `.xterm-char-measure-element` measured 0 while
an identical monospace span in the same container measured ~295px; `max-width: none`
on the measure element restored ~295px, and reopening the terminal with the exemption
active rendered the prompt with `.xterm-screen` sized 369x760. No amount of
remeasure/refit can fix this — the CSS re-caps the measurement to 0 every time — so
the FN-7692 CharSizeService guard is removed.
- Exempt `.xterm-helpers` / `.xterm-char-measure-element` from the mobile max-width
reset in styles.css (covers both TerminalModal and SessionTerminal)
- Revert the ineffective FN-7692 remeasure guard and its tests
- Update changeset (patch) and the docs/solutions write-up to the real root cause
Note: root cause + fix validated in the automation browser via mobile emulation
(393px, iPhone UA, forced touch), not a physical device.
Fusion-Task-Id: FN-7693
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Correct the planner-oversight confirmation messaging so it no longer claims a hard block when the active auto-merge policy will actually advance the merge/pull-request stage unattended.
- decidePlannerRecovery accepts an additive, messaging-only `autoMergeWillProceed` flag and picks accurate reason wording (advisory vs. genuine human-approval block vs. neutral/unknown) for merger/pull-request await_confirmation decisions
- PlannerRecoveryController.tick threads `allowsAutoMergeProcessing(task, settings)` into decidePlannerRecovery as `autoMergeWillProceed`
- project-engine's requestConfirmation steering comment prefix changed from "confirmation required" to neutral "merge checkpoint" so it doesn't contradict the now-accurate reason text
- added regression tests in planner-recovery.test.ts and planner-overseer-intervention-wiring.test.ts
- added changeset and doc note
Files changed:
.changeset/fn-7692-merger-confirmation-copy.md | 7 +++
docs/architecture.md | 10 +++-
packages/core/src/__tests__/planner-recovery.test.ts | 66 ++++++++++++++++++++++
packages/core/src/planner-recovery.ts | 36 +++++++++++-
packages/engine/src/__tests__/planner-overseer-intervention-wiring.test.ts | 37 ++++++++++++
packages/engine/src/planner-recovery-controller.ts | 14 ++++-
packages/engine/src/project-engine.ts | 11 +++-
7 files changed, 176 insertions(+), 5 deletions(-)
Fusion-Task-Id: FN-7692
Fusion-Task-Lineage: 187684b8-1d24-425d-85d4-627587469908
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
## Summary
`RuntimeFallbackBadge.tsx` (added by an earlier change) calls
`useToast()` unconditionally. It's rendered inside `LiveAgentCard` (used
by `ActiveAgentsPanel`) and on both the board/list agent cards in
`AgentsView`, but the render helpers in `ActiveAgentsPanel.test.tsx` and
`AgentsView.test.tsx` never wrapped the component under test in a
`ToastProvider`. Every test that mounts a card threw `"useToast must be
used within ToastProvider"`.
## Fix
Adds a `renderPanel()` / `renderView()` helper to each test file —
matching the exact wrapping pattern already used correctly in
`RuntimeFallbackBadge.test.tsx` — and routes every `render(...)` call
site through it. No test assertions, fixtures, or expected behavior were
changed; this is strictly a render-setup fix.
## Before / After
- `ActiveAgentsPanel.test.tsx`: 12/15 → 15/15 passing
- `AgentsView.test.tsx`: ~96-108/132 → 130/132 passing (the 2 remaining
failures are pre-existing, unrelated CSS-content assertion failures —
not `useToast`-related — and are out of scope for this fix)
- `RuntimeFallbackBadge.test.tsx`: unaffected, still 8/8 passing
## Verification
- `pnpm --filter @fusion/dashboard exec vitest run
app/components/__tests__/ActiveAgentsPanel.test.tsx` — 15/15 pass
- `pnpm --filter @fusion/dashboard exec vitest run
app/components/__tests__/AgentsView.test.tsx` — 130/132 pass
- `pnpm --filter @fusion/dashboard exec vitest run
app/components/__tests__/RuntimeFallbackBadge.test.tsx` — 8/8 pass
- `pnpm --filter @fusion/dashboard run typecheck` — clean
- `eslint` on both modified files — clean
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Tests**
* Updated dashboard component tests to render within the toast context,
preventing crashes when nested UI needs toast access.
* Standardized test helpers for the agents views to better match real
app behavior.
* Adjusted one token-usage assertion to open the controls popup before
checking its content.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
A Staff Engineer pre-landing review (Greptile/CodeRabbit) on #1957
(merged) flagged four structural issues. This PR fixes the three that
were confirmed still present on `main`; the fourth (an unregistered-rule
`eslint-disable-next-line react-hooks/exhaustive-deps` comment) was
already fixed in #1957's second commit before merge and needed no
further change.
1. **`resolvePluginRuntime()` mislabeled "found but failed to init" as
`not_found`.** When a `runtimeHint` plugin registration is found but
`pluginContext`/`createRuntimeContext(...)` comes back falsy, the
resolver returned `reason: "not_found"` — indistinguishable from "never
registered" — defeating the point of a distinct `FallbackReason`. Now
returns `reason: "init_error"`. Updated the existing test that wrongly
asserted `"not_found"` for this path, and added a new test asserting all
three reachable `FallbackReason` values (`not_found`, `init_error`,
`factory_error`) are pairwise distinct.
2. **`ActiveAgentsPanel.tsx`/`AgentsView.tsx` hardcoded
`isInViewport={true}`.** Every agent card (live-agent header, board
card, list card) polled the runtime-fallback endpoint every 30s forever,
even scrolled off-screen — unlike `TaskCard.tsx`'s correct
`IntersectionObserver`-gated pattern. Both files now thread a real
`IntersectionObserver`-backed viewport signal into
`RuntimeFallbackBadge`. Added regression tests proving polling stops
once a badge instance's `isInViewport` transitions to `false` and
resumes once it goes back to `true` (desktop + a mobile-breakpoint
variant), plus verified via `tsc --noEmit` for `@fusion/dashboard`.
3. **Toast dedupe was per-hook-instance, not shared.**
`useRuntimeFallbackStatus`'s `lastToastedEventIdRef` was a local
`useRef`, so the same task rendered simultaneously in two card surfaces
(e.g. `ActiveAgentsPanel` + `AgentsView`) fired two separate toasts for
one fallback event. Dedupe now lives in module-level shared state (a
bounded `Map` keyed by `taskId:eventId`, FIFO-evicted past 500 entries)
so a fallback event toasts exactly once across every
simultaneously-mounted badge instance for the same task. Added a
cross-instance regression test mounting two badges for the same
`taskId`/`eventId` and asserting exactly one toast fires.
## Test evidence
- `pnpm --filter @fusion/engine exec vitest run
src/__tests__/runtime-resolution.test.ts --reporter=dot` — 25/25 pass
- `pnpm --filter @fusion/dashboard exec vitest run
app/components/__tests__/RuntimeFallbackBadge.test.tsx --reporter=dot` —
11/11 pass
- `pnpm --filter @fusion/dashboard run typecheck` — clean
- `pnpm --filter @fusion/engine run typecheck` — clean
## Scope
Isolated 6-file diff on top of current `main`
(`packages/engine/src/runtime-resolution.ts`,
`packages/engine/src/__tests__/runtime-resolution.test.ts`,
`packages/dashboard/app/hooks/useRuntimeFallbackStatus.ts`,
`packages/dashboard/app/components/ActiveAgentsPanel.tsx`,
`packages/dashboard/app/components/AgentsView.tsx`,
`packages/dashboard/app/components/__tests__/RuntimeFallbackBadge.test.tsx`).
No behavior outside the three findings above was touched.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **New Features**
- The desktop dashboard now supports plugin-backed runtime features,
improving how plugin-enabled workflows are loaded and run.
- Agent cards now pause background fallback polling when they’re
off-screen, helping the dashboard feel smoother and more responsive.
- **Bug Fixes**
- Improved runtime fallback handling so missing runtimes and
initialization failures are reported more accurately.
- Toast notifications are now better deduplicated, reducing repeated
alerts when multiple views show the same fallback state.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
RuntimeFallbackBadge.tsx calls useToast() unconditionally, but the
ActiveAgentsPanel.test.tsx and AgentsView.test.tsx render helpers never
wrapped the component under test in a ToastProvider, causing every mount
that transitively renders RuntimeFallbackBadge to throw
'useToast must be used within ToastProvider'.
Adds renderPanel()/renderView() helpers (matching the existing
RuntimeFallbackBadge.test.tsx pattern) that wrap render(...) calls in
ToastProvider, and routes every render call site in both files through
them. No test assertions or fixtures were changed.
Fixes 12/15 ActiveAgentsPanel.test.tsx failures and ~96-108/132
AgentsView.test.tsx failures.
The mobile terminal rendered blank even though the WebSocket was Connected and
the shell prompt had already streamed in. Root cause (reproduced live): on the
mobile fullscreen layout xterm's CharSizeService can measure a 0-width character
cell, so FitAddon.fit() proposes 0 columns/rows and .xterm-screen (plus the WebGL
canvas) collapses to 0x0 — prompt bytes arrive and are written into xterm's row
DOM but paint into a zero-size box. Renderer-independent and mobile-layout-
specific; not fixed by resize/font-size re-fits because prior guards only validate
the container width and font load, never the resulting measured screen/cell width.
Add guardAgainstCollapsedTerminalScreen (app/utils/terminalPreferences.ts) and arm
it from both terminal surfaces (TerminalModal + SessionTerminal) right after their
initial fit. While the container has a width but .xterm-screen does not, it forces
a genuine DOM-strategy remeasure (forceTerminalFontRemeasure) + fit, re-driven by a
ResizeObserver until the screen has a real width. It waits (does not give up) while
the container is not yet measurable, is bounded so it never spins, and is disposed
on every re-init/close path. Recurrence of FN-7620/FN-7686.
- Add isTerminalScreenCollapsed + guardAgainstCollapsedTerminalScreen with tests
- Wire + dispose the guard across all xterm (re)init/close paths in both surfaces
- Add changeset (patch) and a docs/solutions write-up
Note: reproduced via mobile emulation (393px, iPhone UA, forced touch), not a
physical device; the guard is the structural fix — confirm on a real device.
Fusion-Task-Id: FN-7692
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Custom providers previously never enabled prompt-cache control, so agent turns re-billed the full context every request even on cache-capable backends.
- Add `CustomProvider.anthropicPromptCaching` opt-in flag in @fusion/core types
- Set pi-ai `compat.cacheControlFormat="anthropic"` on opted-in models in both registration paths: custom-provider-registry `toProviderConfig` and pi.ts `createFnAgent`
- Expose the new toggle in the dashboard CustomProvidersSection UI (with supporting CSS) and thread it through the legacy API + custom-provider routes
- Update docs (dashboard-guide, settings-reference) to document the new setting
- Add engine test coverage for the caching flag across provider registration and pi-create-fn-agent paths
- Add changeset for the fix
Files changed:
.changeset/fn-7689-custom-provider-prompt-caching.md | 7 +
docs/dashboard-guide.md | 1 +
docs/settings-reference.md | 2 +-
packages/core/src/types.ts | 15 ++
packages/dashboard/app/api/legacy.ts | 12 ++
packages/dashboard/app/components/CustomProvidersSection.css | 24 +++
packages/dashboard/app/components/CustomProvidersSection.tsx | 58 +++++-
packages/dashboard/src/routes/register-custom-provider-routes.ts | 16 ++
packages/engine/src/__tests__/pi-create-fn-agent.test.ts | 71 +++++++
packages/engine/src/__tests__/provider-registration.test.ts | 204 ++++++++++++++++++++-
packages/engine/src/custom-provider-registry.ts | 71 +++++--
packages/engine/src/pi.ts | 27 ++-
12 files changed, 473 insertions(+), 35 deletions(-)
Fusion-Task-Id: FN-7689
Fusion-Task-Lineage: b4f88f32-50da-4651-a546-432a95a1ab1c
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Investigated whether --login in TerminalService first-prompt latency is a meaningful contributor and added a one-time diagnostic hint plus documentation of findings.
- Add SLOW_LOGIN_PROFILE_HINT_MS (2000ms) threshold and one-time, non-blocking console.info hint in createSession()'s PTY onData handler when a login shell is slow to produce first output
- Track spawnStartedAt and loginProfileHintLogged per session, and whether the succeeding spawn attempt used --login, without altering spawn args, timeouts, or the retry-without-login fallback
- Add regression tests covering the slow-login-profile hint behavior in terminal-service.test.ts
- Document the investigation and findings in docs/solutions/developer-experience/login-shell-profile-latency.md and link it from docs/dashboard-guide.md
- Add a patch changeset for @runfusion/fusion describing the new server-log hint
Files changed:
.changeset/fn-7688-login-shell-profile-latency.md | 7 ++
docs/dashboard-guide.md | 17 +++
.../login-shell-profile-latency.md | 79 ++++++++++++++
.../src/__tests__/terminal-service.test.ts | 121 +++++++++++++++++++++
packages/dashboard/src/terminal-service.ts | 57 ++++++++++
5 files changed, 281 insertions(+)
Fusion-Task-Id: FN-7688
Fusion-Task-Lineage: 08d5dd47-ea9f-4973-9f0e-a8d5fdeae111
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
On tablet-width viewports (769-1024px) the terminal header was overcrowded, so shortcut/zoom/preference controls now move into the bottom status-bar footer instead, mirroring the existing FN-7560 mobile layout while true desktop keeps the header controls.
- Add an isTabletTerminal detection flag (769-1024px, non-mobile) alongside the existing mobile flag
- Render the shared terminalActionControls fragment in the .terminal-status-bar footer for tablet widths, keeping desktop pin/pop-out toggles available there too
- True desktop (>1024px) continues to render font size / clear / shortcuts / preferences controls in the header
- Update TerminalModal.css with footer layout styles for the tablet action controls
- Update TerminalModal tests to cover the new tablet breakpoint rendering
- Update docs/dashboard-guide.md to describe the tablet footer control location
- Add a changeset (@runfusion/fusion: patch) documenting the fix
Files changed:
.changeset/fn-7684-tablet-terminal-footer.md | 7 ++
docs/dashboard-guide.md | 6 +-
.../dashboard/app/components/TerminalModal.css | 34 ++++++
.../dashboard/app/components/TerminalModal.tsx | 103 ++++++++++++-----
.../components/__tests__/TerminalModal.test.tsx | 126 ++++++++++++++++++---
5 files changed, 231 insertions(+), 45 deletions(-)
Fusion-Task-Id: FN-7684
Fusion-Task-Lineage: 627c40af-a152-451a-a045-65ad8babea9e
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
