## Summary
Main Full Suite shards have been red after recent landings. Root causes:
1. **Executor tests** — `execute()` now polls
`getTaskVerificationRequestAsync` (chat-enqueued verification). Shared
`createMockStore()` (and soft-delete inline store) lacked the method, so
nearly every execute-path suite failed with `is not a function`.
2. **TaskDetailModal suites** — `NativeStructurePreview` imports `Map` /
`Lightbulb` / `BarChart3` / `Target` / `CircleAlert` from lucide; the
shared TaskDetail lucide mock omitted them, so suites failed at import.
3. **Grok process-lifecycle** — 15s bound stress timed out under
full-suite load without product-bug evidence → quarantined on sight per
AGENTS.md.
## Test plan
- [x] `executor-task-done-blocked`, `executor-fast-mode-workflows`,
concurrent-execute race
- [x] `executor-step-session`, plan-only scope leak, review-step
indexing
- [x] `TaskDetailModal.create-pr` + `TaskDetail.mobile-transition`
- [ ] Full Suite CI on this PR
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Improvements**
* Added `html2canvas` support in the dashboard to enable HTML-to-canvas
rendering needed for visual structure previews.
* **Tests**
* Updated task execution test mocks to handle task verification-request
flows reliably.
* Improved task deletion safeguard coverage and related execution
behavior checks.
* Enhanced test stubs to support structure preview rendering elements
during modal-related tests.
* **Chores**
* Quarantined a timing-sensitive process lifecycle test and refreshed
quarantine tracking to improve full-suite stability.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
Full Suite after #2288 still red on shard 4: `process-lifecycle.test.ts`
times out at 5s under shard transform load even after reducing
reimports.
- Move Symbol.for `process.exit` reaper onto `process-manager`
(lifecycle owner)
- Stress test reimports that module (not the full plugin graph)
- Explicit 15s timeout for the cold-transform bound stress test
## Test plan
- [x] Local process-lifecycle green (~2s)
- [ ] PR merge gate
- [ ] Post-merge Full Suite green
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **Bug Fixes**
- Improved Grok runtime process cleanup during application shutdown.
- Prevented duplicate cleanup handlers from accumulating during repeated
module loading.
- Ensured managed processes are reliably terminated when the process
exits.
- **Tests**
- Expanded lifecycle coverage to validate repeated loading scenarios.
- Increased test timeouts for more reliable stress-test execution.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
#2287 cleared most of Full Suite run 29633869887 / post-merge
29634793317. Shards 1 and 2 went green; remaining floaters:
- **TaskDetailModal.css FN-8154** — assert FN-8166 zeroed mobile
`.detail-activity` inset (`padding-inline-end: 0`), not the pre-8166
`var(--space-md)` residual
- **QuickEntryBox** — wait for cleared input / priority button after
create under shard load (not only `onCreate` mock call)
- **grok process-lifecycle** — prove Symbol.for exit-hook bound with 2
reimports so transform cost stays under 5s on full-suite shards
## Test plan
- [x] Local: FN-8154, QuickEntry clear/priority, process-lifecycle
- [ ] PR merge gate green
- [ ] Post-merge Full Suite on `main` green
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Tests**
* Improved Quick Entry tests to wait for textarea clearing and verify
priority resets to normal after task creation.
* Updated CSS contract assertions for mobile task-detail activity
spacing to match the latest overlay/inset behavior.
* Streamlined the Grok plugin process lifecycle test loop to run fewer
module-evaluation iterations while keeping existing lifecycle and
warning checks.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
Restacks onto latest main after #2285 and clears the remaining Full
Suite red classes from run
[29633869887](https://github.com/Runfusion/Fusion/actions/runs/29633869887)
(post-#2285):
- **Shard 1:** `agent-skills-flow` vitest TDZ — hoist `mockFiles` via
`vi.hoisted` (same class as skill-resolver in #2285)
- **Shard 2/3:** incomplete mocks after product drift
- `isFullScreenSheetViewport` / `isShortViewport` on viewport mocks
(without overriding dynamic mobile helpers)
- `fetchCodebaseMetrics` on Command Center `api/legacy` mocks
- `fetchSettings` on `agent-modals-mobile` api mock
- **Shard 3:** PlanningMode `ui-interactions` race — sync-settle
`fetchGlobalSettings` (FN-8245 pattern from planning-flow)
- **Shard 3:** settings search drift guard — inventory
`SettingsFieldRow` `htmlFor` keys (`mobileNavPrimaryItems`)
- **Shard 2:** FloatingWindow shared-stack product bug — only reclaim
z-index on hidden→visible (not every mount effect), so last-mounted
utility stays on top
- **Shard 4:** grok process-lifecycle timeout under shard load — prove
bound with 5 reimports instead of 15
## Test plan
- [x] `agent-skills-flow.test.ts` green
- [x] `process-lifecycle.test.ts` green
- [x] FileBrowserModal, FloatingWindowStack.cross-type,
agent-modals-mobile, settings-search-index, SystemControlsArea,
PlanningModeModal.ui-interactions + planning-flow (210 tests) green
- [ ] PR merge gate (Lint/Typecheck/Build/Gate)
- [ ] Post-merge Full Suite on `main` green
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved floating-window stacking so reopened or interacted windows
appear in the correct order.
* Restored consistent layering between floating windows and expanded
dock modals.
* **Tests**
* Updated automated coverage for viewport behavior, codebase metrics,
settings search indexing, and process lifecycle scenarios.
* Improved test reliability and consistency across responsive layouts
and modal interactions.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What happened
FN-8004's implementation work finished and passed review. The auto-merge
then failed with `Grok ACP turn failed: Internal error` — a ~20 second
provider blip — and the task was parked `status: "failed"` with 8 files
of complete, reviewed work stranded on its branch.
The park is the interesting part: `status: "failed"` is precisely what
tells recovery to stop. So a misclassification here isn't a missed
retry, it's **terminal**. Both recovery paths were disabled by the same
wrong verdict:
- `maybeRetryTransientMerge` (inline, 3 retries w/ backoff) — never
fired once (`mergeTransientRetryCount: 0`).
- `recoverTransientMergeFailures` (self-healing sweep, exists exactly to
rescue parked in-review tasks) — skipped it, gated on the same
classifier.
## Three defects fixed
**1. No AI-provider failure class existed.** The AI merge drives a real
LLM turn, but `classifyTransientMergeError` only modeled git/lease/spawn
faults. Adds `ai-provider-turn-failure`.
**2. ACP dropped the error detail.** `promptAcpSession` rethrew the SDK
error unchanged, discarding the JSON-RPC `code`/`data` — the only
evidence the fault was provider-side. ("Internal error" is just the
standard text for `-32603`.) It now preserves them, keeping the original
as `cause`:
```
Internal error (acp rpc code -32603, retryable)
```
Classification anchors on that envelope, **not** on the bare `"Internal
error"` — matching that unanchored would disguise genuine application
defects as retryable blips. Only provider-fault codes (`-32603`,
`-32000`..`-32003`) are retryable; caller-fault codes
(`-32600`..`-32602`) stay permanent, since retrying just repeats the
failing call.
**3. Sweep/inline asymmetry** (found while tracing; latent and
unreported). The inline gate accepted `isTransientError(msg) ||
classify(msg)`, but the sweep consulted **only** the classifier. So
`ECONNRESET` / `socket hang up` during a merge earned inline retries and
then went **invisible to the sweep** once parked — stranded forever. The
classifier now delegates to `isTransientError`, so both gates agree by
construction.
To keep that delegation from importing the detector's
`usage-limit-detector → logger` chain (the chain FN-5627 split the
classifier out to avoid, which would break
`notification-service.test.ts`'s partial `vi.mock`), the pure predicates
moved to the import-free leaf `transient-error-patterns.ts`, re-exported
from `transient-error-detector.ts`. All 13 exports preserved, verified
programmatically.
## Loosened budgets
Per request, so more self-heals. Both apply **only** to errors already
proven transient; the ceiling and
`merger:transient-failure-budget-exhausted` audit path remain.
| Budget | Before | After |
|---|---|---|
| `MAX_AUTO_MERGE_TRANSIENT_RETRIES` | 3 | 5 (backoff
5s/10s/20s/40s/80s) |
| `MAX_TRANSIENT_MERGE_RECOVERIES` | 2 | 5 |
The bump broke two suites that had hardcoded the old `3`. Rather than
swap in another magic number, both now derive the cap from the constant
so future tuning doesn't re-break them.
## Verification
- `pnpm test:gate` green · `pnpm lint` clean · engine + ACP typecheck
clean · `pnpm verify:fast` PASS (5/5)
- ACP plugin 230 tests green · Grok plugin 64 green · engine
transient/merge suites 136 green
- Regression tests assert the **invariant across every surface** (per
*Fix the Invariant, Not the Repro*), not just the reported Grok string:
both ACP runtime prefixes, all retryable/non-retryable rpc codes, both
SDK error shapes, network delegation, class-ordering, and negative cases
proving bare `"Internal error"` and real defects stay permanent.
- A test caught a genuine bug in my own code mid-review (nested-shape
message shadowing), now fixed.
- `notifier.test.ts > "awaiting approval"` fails — **confirmed
pre-existing on clean main**, unrelated.
## Note
FN-8004's own branch (`fusion/fn-8004`) is still unmerged and its work
looks complete. Once this lands, its merge should be retried separately.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Grok advertises promptCapabilities.image=false and ignores ACP image
ContentBlocks (live probe: NO_IMAGE). Path-based vision works when the
agent is given an absolute file path. Include path hints in chat prompts
from .fusion/chat-attachments and carry path on ChatImageContent for
file:// uris.
AcpRuntimeAdapter.promptWithFallback ignored options, so dashboard chat
images never became ACP ContentBlock image entries. Extract images from
prompt options and pass them through buildPromptBlocks for both acp-runtime
and the Grok vendored client.
Grok emits `_x.ai/session_notification` / `_x.ai/session/update` for
hook_execution status. The ACP SDK routes those to Client.extNotification;
without it, every successful post_tool_use hook logged -32601 Method not
found. Implement no-op extMethod/extNotification on default and bridging
handlers in acp-runtime and the Grok vendored copy.
Replace one-shot grok -p JSON with native grok agent stdio (ACP) for realtime
streaming, tool visibility, and multi-turn sessions. Vendor the ACP client
into fusion-plugin-grok-runtime, forward Fusion fn_* tools and operator MCP,
stage Fusion skills via --plugin-dir, authenticate per xAI headless docs, and
align project chat manager store resolution so Grok chat sessions can send.
Raises Grok and Droid CLI cold-start timeout defaults from 60s to 120s and lets operators override them via environment variables.
- Grok runtime adapter: new GROK_CLI_FIRST_OUTPUT_TIMEOUT_MS env override for the first-output cold-start guard; default raised 60000ms → 120000ms; invalid/non-positive values fall back to the default
- Droid provider: new PI_DROID_CLI_FIRST_LINE_TIMEOUT_MS env override for the first-line cold-start guard; default raised 60000ms → 120000ms; invalid/non-positive values fall back to the default
- 30-minute inactivity safety net left unchanged on both adapters
- Added/updated unit tests covering the new env-driven timeout resolution and fallback behavior for both plugins
- Documented the new settings in docs/settings-reference.md and both plugin READMEs
- Added a minor changeset for @runfusion/fusion
Files changed:
.changeset/fn-7838-cli-timeout-configurable.md | 7 ++
docs/settings-reference.md | 13 ++-
plugins/fusion-plugin-droid-runtime/README.md | 8 ++
.../src/__tests__/provider.test.ts | 97 +++++++++++++++++++++-
.../fusion-plugin-droid-runtime/src/provider.ts | 26 ++++--
plugins/fusion-plugin-grok-runtime/README.md | 8 ++
.../src/__tests__/runtime-adapter.test.ts | 52 +++++++++++-
.../src/runtime-adapter.ts | 23 ++++-
8 files changed, 222 insertions(+), 12 deletions(-)
Fusion-Task-Id: FN-7838
Fusion-Task-Lineage: 7aad4470-b28f-42bd-be01-8363a1dd05e5
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
The running dashboard loads the compiled plugin dist, which was stale and
invoked `grok --prompt <text> --format json --directory <cwd>` — flags grok
0.2.93 rejects ("unexpected argument '--prompt'"), yielding a non-zero exit,
no JSON, and an empty "No message" bubble. The source already switched to the
valid `grok -p <text> --output-format json [-m <model>] [--cwd <dir>]`
contract (FN-7790/FN-7796); this rebuilds dist to match.
Also reconcile the FN-7779 test suite: a genuinely empty response is a parsed
`{text:"",stopReason:"EndTurn"}` object, not zero stdout bytes, so the
"stays silent" test now models that shape instead of contradicting the
FN-7796 zero-stdout wrong-binary diagnostic. All 64 plugin tests pass.
Fusion-Task-Id: FN-7779
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Narrative: the streaming-json headless contract intermittently emitted only thought events then stopReason:Cancelled with zero text, leaving Chat replies silently empty; the adapter now spawns grok with --output-format json, buffers stdout, and parses the single JSON response on process close, with streaming-json parsing kept only as a diagnostic fallback.
- Change grok CLI invocation from --output-format streaming-json to --output-format json (cli-stream.ts)
- Add GrokCliJsonResponse type ({text, stopReason, sessionId, requestId, thought}) and parseJsonOutput() to stream-parser.ts, keeping legacy NDJSON line parsing for fallback/diagnostics
- Rework runtime-adapter.ts to buffer full stdout, parse it via parsePromptOutput (JSON object first, NDJSON fallback), and surface a formatTerminalNoTextDiagnostic when a non-EndTurn stopReason yields no assistant text
- Rename first-line/inactivity timeout bookkeeping from line-based to output/chunk-based (FIRST_OUTPUT_TIMEOUT_MS, firstOutputReceived, firstStdoutChunk) since stdout is no longer consumed via readline
- Update cli-stream/runtime-adapter/stream-parser tests to cover the JSON response path and the Cancelled/no-text diagnostic
- Update docs/grok-cli-contract.md and plugin README to document the json output-format contract and diagnostics
- Add changeset fn-7796-grok-cli-reliable-headless.md (patch, fix)
Files changed:
.changeset/fn-7796-grok-cli-reliable-headless.md | 7 +
docs/grok-cli-contract.md | 108 ++++++++-----
plugins/fusion-plugin-grok-runtime/README.md | 16 +-
.../src/__tests__/cli-stream.test.ts | 4 +-
.../src/__tests__/runtime-adapter.test.ts | 73 ++++++++-
.../src/__tests__/stream-parser.test.ts | 80 +++++----
.../fusion-plugin-grok-runtime/src/cli-stream.ts | 14 +-
.../src/runtime-adapter.ts | 180 ++++++++++++---------
.../src/stream-parser.ts | 65 ++++++--
plugins/fusion-plugin-grok-runtime/src/types.ts | 13 +-
10 files changed, 373 insertions(+), 187 deletions(-)
Fusion-Task-Id: FN-7796
Fusion-Task-Lineage: c920fcf0-98f8-42ec-867a-7f76c0aca1b7
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Align the Grok CLI runtime plugin adapter with the real xAI grok CLI stream contract so tool responses are no longer silently dropped, and update tests/docs to match.
- Rework stream-parser.ts to parse the actual grok CLI event/message shape instead of the previously assumed schema
- Trim runtime-adapter.ts and types.ts down to the fields the real CLI contract emits, removing speculative/unsupported fields
- Update cli-stream.ts to match the corrected event handling
- Rewrite runtime-adapter, stream-parser, and cli-stream test suites to exercise the real CLI contract end-to-end
- Update docs/grok-cli-contract.md and plugin README to document the verified contract
- Add changeset for the grok-runtime plugin fix
Files changed:
.changeset/fn-7790-grok-cli-real-contract.md | 7 +
docs/grok-cli-contract.md | 432 +++++++--------------
plugins/fusion-plugin-grok-runtime/README.md | 149 ++-----
.../src/__tests__/cli-stream.test.ts | 22 +-
.../src/__tests__/runtime-adapter.test.ts | 250 ++++--------
.../src/__tests__/stream-parser.test.ts | 108 ++----
.../fusion-plugin-grok-runtime/src/cli-stream.ts | 22 +-
.../src/runtime-adapter.ts | 109 ++----
.../src/stream-parser.ts | 27 +-
plugins/fusion-plugin-grok-runtime/src/types.ts | 104 +----
10 files changed, 338 insertions(+), 892 deletions(-)
Fusion-Task-Id: FN-7790
Fusion-Task-Lineage: 377e86c4-c005-46a5-9eb4-69e858c38b79
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Fixes the residual "Grok CLI still returns no message immediately" case where a headless run exits code 0 but never emits any parsed NDJSON event, previously treated as a silent success.
- Detect a code-0 close with zero parsed NDJSON events and surface a diagnostic explaining the likely cause (wrong/unsupported grok binary falling into interactive mode and hitting EOF on stdin).
- Track and emit assistant text/diagnostics via a new appendMessage/emitDiagnosticText path so onText and session.state.errorMessage stay in sync, including on spawn failure and inactivity/first-line timeouts.
- Add first-line/inactivity timeout diagnostics with concrete elapsed-time messaging instead of silent kills.
- Add regression coverage in runtime-adapter.test.ts and grok-runtime-routing.test.ts for the zero-NDJSON exit path.
- Document the contract update in docs/grok-cli-contract.md.
- Add a patch changeset for @runfusion/fusion.
Files changed:
$(cat /tmp/diffstat_fn7788.txt)
Fusion-Task-Id: FN-7788
Fusion-Task-Lineage: dbb238a9-9601-47fc-8a88-40817d749337
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Root-causes the Grok CLI empty-response bug: readline close no longer finalizes the session before the subprocess close event can attach exit code/stderr diagnostics, so failures were silently swallowed as empty assistant messages.
- Wait for subprocess close/error (not readline close) to finalize the Grok CLI session, so non-zero exits can attach stderr before callers inspect the result
- Add GrokSession.state.errorMessage to carry concrete diagnostics (spawn failure, process error, non-zero exit + stderr, or NDJSON error event) through the resolve-never-reject runtime contract
- Track whether any text was received so error diagnostics are only recorded when the run actually produced nothing
- Add a changeset documenting the fix for @runfusion/fusion
- Extend runtime-adapter tests to cover spawn failure, process error, non-zero exit with/without stderr, and NDJSON error-event diagnostics
Files changed:
.changeset/fn-7782-grok-cli-no-response.md | 7 ++
.../src/__tests__/runtime-adapter.test.ts | 98 ++++++++++++++++++++--
.../src/runtime-adapter.ts | 75 +++++++++++++----
plugins/fusion-plugin-grok-runtime/src/types.ts | 1 +
4 files changed, 161 insertions(+), 20 deletions(-)
Fusion-Task-Id: FN-7782
Fusion-Task-Lineage: c2907a70-0556-488f-bda3-132657b64071
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
