- decideExecutionPlan: implicit wide-blast reasons (missing base, diff failed,
no changes, shared infra, unmapped package) route to new gate mode instead of full
- CI===true force-full branch removed (CI no longer calls test-changed.mjs)
- changed mode runs pnpm test:gate before the affected set
- full suite reachable only via explicit --full / FUSION_TEST_FULL=1
- characterization tests updated + new only-explicit-full invariant test
- pr-checks.yml: gate job = boot smoke + pnpm test:gate; shards/slow/inventory removed
- full-suite.yml (new): 4-way shards, engine slow tier, inventory guard on push to main
- ci.yml deleted (trigger-disabled since FN-1541; was dead config)
- ci-workflow.test.ts rewritten to pin the new gate shape; release/signing blocks preserved
- docs/contributing.md: pnpm test:gate is the merge gate; verify:workspace repositioned as deep opt-in
BRANCH-PROTECTION CUTOVER: set required checks to exactly [Lint, Typecheck, Build, Gate] after merging; open PRs must rebase.
The merge of main added resolvePrBaseRef (consumes tryRun mocks before the
push sequence) and classifyGhError mapping (network errors -> structured
502 with githubError payload). The long-orphaned test pinned the old shape.
Re-mapped mock queues to the real sequence, asserted the stronger current
error contract (status + githubError.code + retryable), and made the
accidentally-passing no-commits test deterministic. Mutate-to-prove
verified the push assertion still bites.
The trial's pnpm add+remove cycle left peer-suffix entries inconsistent
(vitest@3.2.4 combo missing); local warm installs validated shallowly while
CI's fresh resolution failed every job. Re-resolved with no dependency
changes (package.json untouched).
vi.waitFor polls, so the bound adds zero time to green runs; ink frame
scheduling has flaked past 3s under shard CPU contention while passing
instantly in isolation.
- WorkflowNodeEditor overlay was missing the `open` class, so the
graph editor mounted with display:none — clicking the button just
dismissed the steps view. Add `open` so the overlay renders.
- Add fusion-plugin-compound-engineering and fusion-plugin-roadmap to
BUILTIN_PLUGINS so they show under Settings → Built-in Plugins.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Show a one-time GitHub star prompt after a task first reaches done.
- detect task status transitions into done and trigger the prompt only in project view
- add a dismissible GitHub star prompt component plus localStorage-backed persistence hook
- cover the new prompt behavior with component, hook, and transition helper tests
- document the prompt behavior and styling guidance in the dashboard guide
Files changed:
docs/dashboard-guide.md | 3 +
packages/dashboard/app/App.tsx | 21 +++++-
.../dashboard/app/components/GitHubStarPrompt.css | 76 ++++++++++++++++++++++
.../dashboard/app/components/GitHubStarPrompt.tsx | 53 +++++++++++++++
.../app/components/__tests__/App.test.tsx | 12 +++-
.../components/__tests__/GitHubStarPrompt.test.tsx | 40 ++++++++++++
.../hooks/__tests__/useGitHubStarPrompt.test.ts | 64 ++++++++++++++++++
.../dashboard/app/hooks/useGitHubStarPrompt.ts | 46 +++++++++++++
8 files changed, 313 insertions(+), 2 deletions(-)
Fusion-Task-Id: FN-5967
Fusion-Task-Lineage: 5fffdf4d-61d8-4ac8-bcf4-8b453b639b28
Make the Create Pull Request modal interactive before AI metadata finishes loading.
- load preflight checks and PR options independently from AI title/body generation with separate loading and error states
- preserve base-branch selection and remediation actions while refreshing preflight or regenerating metadata
- cover the faster modal flow in tests and document the immediate-render behavior in the dashboard guide
Files changed:
docs/dashboard-guide.md | 3 +-
packages/dashboard/app/components/PrCreateModal.css | 11 +
packages/dashboard/app/components/PrCreateModal.tsx | 357 ++++++++++++++-------
packages/dashboard/app/components/__tests__/PrCreateModal.test.tsx | 110 ++++++-
4 files changed, 355 insertions(+), 126 deletions(-)
Fusion-Task-Id: FN-5954
Fusion-Task-Lineage: 2595d220-ca7a-403c-935c-900034a20c67
Restore the chat message pane empty state on mobile without regressing desktop sidebar layouts.
- style chat message-pane empty states as centered framed cards and normalize mobile spacing tokens
- keep sidebar padded empty states left-aligned so loading and list placeholders preserve their desktop layout
- add mobile and desktop regression coverage for direct chats and chat rooms across empty, loading, populated, and streaming states
Files changed:
packages/dashboard/app/components/ChatView.css | 26 +-
.../__tests__/ChatView.mobile-render.test.tsx | 299 +++++++++++++++++++++
2 files changed, 323 insertions(+), 2 deletions(-)
Fusion-Task-Id: FN-5948
Fusion-Task-Lineage: f4165f04-363d-4ade-8c10-6d242f943c6c
Harden executor worktree recovery coverage against fixed-path collisions.
- create a unique temporary root for the stale unregistered worktree test case
- derive the stale worktree path from that temporary root instead of a shared /tmp/test path
- initialize TaskExecutor with the same temporary root so setup and assertions stay aligned
Files changed:
packages/engine/src/__tests__/executor-worktree.test.ts | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
Fusion-Task-Id: FN-5965
Fusion-Task-Lineage: 4436bd02-b7f7-4e62-8514-0aef6cb67813
Add in-app preflight remediation so Create PR can push task branches without leaving Fusion.
- add a dashboard API route and client helper to push the task branch to origin and recompute PR preflight state
- update the Create Pull Request modal, styles, and tests to surface push-branch remediation alongside AI conflict resolution
- document the flow and add a published CLI changeset plus server coverage for the new push-branch endpoint
Files changed:
.changeset/fn-5950-pr-push-branch.md | 5 +
docs/dashboard-guide.md | 2 +-
packages/dashboard/app/api/legacy.ts | 19 +++
packages/dashboard/app/components/PrCreateModal.css | 4 +-
packages/dashboard/app/components/PrCreateModal.tsx | 48 +++++-
packages/dashboard/app/components/__tests__/PrCreateModal.test.tsx | 44 ++++++
packages/dashboard/src/__tests__/register-git-github.pr-push-branch.test.ts | 176 +++++++++++++++++++++
packages/dashboard/src/routes/register-git-github.ts | 68 ++++++++
8 files changed, 362 insertions(+), 4 deletions(-)
Fusion-Task-Id: FN-5950
Fusion-Task-Lineage: 5a6c5c6a-8f99-44c0-8f54-f0a6ff537ed0
ko now covers every key in every namespace (common/app/errors/cli),
machine-drafted with placeholder/markup preservation. CLI bundles and
the dashboard locale tree regenerated; production build emits per-locale
chunks for all 6 locales with no main-bundle leak (assert passed).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Every shard + the curated-guard job paid ~71s rebuilding 8 packages' dist
from scratch. actions/cache now restores dist on exact content-hash match
(--print-source-hash; branch-switch stable, pure git-based), with a
--seed-artifact-cache step on cache-hit that defeats the mtime trap
(restored dist looks older than checkout-time src mtimes). No restore-keys
partial fallback: stale dist is a known failure mode here. node_modules is
never cached (Windows junction policy). ensure-test-artifacts still runs as
the authority and rebuilds anything genuinely missing or changed.
- db.ts: restrict migration-105 orphan-step cleanup to JSON arrays
(json_type guard so json_each can't expand objects/strings)
- project-engine.ts: requestInterpreterMerge throws on null task lookup
instead of casting null into MergeResult (seam converts to clean failure)
- executor.ts: truncate dual-observe shadow stage walk at the live terminal
stage so healthy in-review tasks don't record a phantom merge transition
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The newly-gated backfill file flaked on a loaded CI shard: waitFor's
private 1s default (independent of the 15s vitest testTimeout) raced the
click->respondToPlanning state-update chain. Passes deterministically in
isolation; 5s bound absorbs shard CPU starvation without masking real
regressions.
Migration 102 defers the destructive agentLogEntries drop until TaskStore
copies legacy rows to JSONL and writes the __meta guard, then relies on a
second init() pass gated on schemaVersion < SCHEMA_VERSION. Migrations
103-105 bump the version to 105 on the first pass, so the second pass never
fired and the legacy table survived forever. Make the drop version-independent
in migrate() and trigger the re-init whenever the legacy table remains.
Also pin secrets-schema.test.ts to String(SCHEMA_VERSION) instead of the
hardcoded "102" string the schema bump invalidated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Components now call useTranslation(); without an initialized instance,
react-i18next's fallback returns inline defaults WITHOUT interpolation
(literal {{count}} in output) and t identity flips once init completes,
double-firing effects that list t in deps. Awaiting a backend-less en
init (all namespaces pre-loaded, useSuspense off) keeps t(key, default,
options) interpolating and identity-stable from the first render.
Cleared ~770 of the 1,051 dashboard test failures; the remainder were
triaged against a clean origin/main worktree baseline: 253 fail
identically there (local-env fake-timer waitFor hangs; CI passes them)
and 23 sweep-caused regressions are being fixed in the round-2 pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- ci-test-shard: timing outputFile is now RELATIVE — one pnpm invocation
fans out to several packages whose vitests all received the same
absolute path, so every package overwrote the same timings file (last
writer wins). Each package now writes <pkgDir>/.timings/; discovery
(discoverWorkspaceTimingFiles) and the CI artifact globs scan the tree
- acp event-bridge-bounds: 20s timeout on the CPU-bound plan-flood test
(timed out at default 5s under loaded CI shard, passes in isolation)
- acp process-manager: port-4040-allowlist marker for its doc comments
(main-side; local guard flagged it after merging main)
Value-audit verdict: the file is already well-factored (~6% genuine bloat).
Merged: 7 full-modal renders for experimental-feature labels into 1;
read-only default-render clusters (Global General defaults, version
display, header actions); export-filename negative folded into positive.
All original assertions survive verbatim; 2 mutate-to-prove checks confirm
merged tests still bite; lane green 3x with no flakes.
Honest note: wall-clock neutral (~47s lane) — per-test cost is dominated
by the ~60-stub beforeEach + full-modal render, not duplicate assertions.
A 30-40% lane reduction requires structural work (shared mock fixture,
subtree renders) deliberately not attempted here (correctness risk).
- gate Quick Chat's session-activation auto-flush to the pre-session
queue only, so a restored queue cannot be sent before the restore
effect's authoritative fetchChatSession check resolves (real flaw —
the original test passed only because the mocked fetch resolved in a
microtask and beat the effect)
- add slow-fetch regression tests in both hooks proving the restored
queue stays un-flushed while server validation is pending
- assert attachChatStream ran before triggering its onDone in the
quick-chat regression test
- drop redundant Promise.resolve() wrapper around fetchChatSession in
useQuickChat's restore effect
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The curated-lane contract still holds — default lane never runs broad
dashboard-app/dashboard-api — but test:quality:api now chains
curated + backfill sub-lanes, so the --project arg lives one level down.
Contract strengthened to also pin the backfill completeness net.
Shard 4 still wedged after the first quarantine — the hang consistently follows
the branch-group fn-001 worktree tests (merge-routing, automerge-precedence,
promotion-gate, pr-sync, single-pr-e2e), with the engine vitest process dying
before printing a summary. These are the suites with known pre-existing
failures (per-task-derived derivation). Move the family to *.slow.test.ts —
the non-required engine-slow lane — alongside the worktree-invariants and
shared-branch-group files. Live-git coverage preserved via test:slow/test:all.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Five reliability-interactions files exercised the same aiMergeTask ->
resolveBranchGroupMergeRouting -> evaluateBranchGroupPromotion triangle,
each paying its own real-git fixture. Removals verified duplicate-by-
duplicate against shared-branch-group-lifecycle (master integration) and
group-merge-coordinator (fast unit coverage of all gate reasons):
- delete branch-group-promotion-gate (all 5 gate scenarios covered elsewhere)
- delete branch-group-promotion (promote-once = lifecycle CASE 3)
- delete shared-group-member-integration; its unique runMaintenance
assertion absorbed into lifecycle CASE 4
- automerge-precedence: drop pause/engine/settings loop (unit-covered);
keep group-level autoMerge override tests (unique)
- merge-routing: drop 2-member + ungrouped routing dups (lifecycle CASEs
2/6); keep worktreePath + dual-audit-event tests (unique)
Inventory diff: 11 removed IDs, all mapping to approved deletions (12th
is the known secrets-crypto randomized-title artifact). Engine
default+reliability: 6516/6516 pass. Affected-file wall-clock 52.9s -> 19.0s.
Skipped: in-process-runtime source-sniff deletion — the audit claim did
not match the actual file (fully-mocked suite, no readFileSync sniffs);
left untouched.
Add docs/solutions/ui-bugs/ learning for the silent animation freeze caused
by transition tokens (duration+easing pairs) used as bare durations, and
update dashboard-guide.md design-token and pitfalls sections to cover the
new --duration-* tokens and the IACVT gotcha.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Document why the FN-5852 queued-message loss survived two fixes (client
gates a destructive flush on a route-level enrichment field that SSE
payloads lack) and seed CONCEPTS.md with Generation, Queued message, and
Enrichment field.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Re-entering a chat flushed the restored queued message based on the
client's stale isGenerating flag (a route-level enrichment the
chat:session:updated SSE payload lacks), firing a send that aborted the
live generation server-side and could lose the message entirely. The
restore path in useChat and useQuickChat now asks the server first:
attach and defer the flush while generating, send immediately only when
no generation is in flight, and keep the bubble on a failed check.
FixesRunfusion/Fusion#1279
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
