- contain fn_artifact_register path payloads: realpath-canonicalized
containment before stat/read — relative paths require and must stay
inside baseDir, absolute paths allowed only under baseDir or the OS
temp dir (deliberate allowance for browser/screenshot tooling);
the process.cwd() fallback is gone, symlink escapes rejected
- bind task-scoped heartbeat artifact registration to the acquired
worktree (baseDir: sessionCwd rebind after acquisition); no-task
heartbeat prompt now says to pass absolute temp-dir paths
- enforce exactly-one payload source (content/uri/dataBase64/path);
content+uri combos are now rejected to match the documented contract
- add FNXC rationale comments at both visual-artifact instruction sites
in the planning prompts (sync contract with the executor prompt)
- media route: statSync -> await stat from node:fs/promises
- range tests ride the in-memory MockSocket harness (TestResponse gains
binary-safe bodyBuffer; real-TCP helper deleted) and assert the full
206 Content-Range/Content-Length contract for every range form
- add PdfViewer coverage (iframe src/title) in DocumentsView tests
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- artifact viewers (image/video lightbox, PDF viewer, doc viewer) now host in
the shared FloatingWindow: draggable by the viewer header, resizable by
edge/corner handles, geometry persisted per viewer kind, closed by button
or Escape; FloatingWindow gains an ariaLabel prop so headerless windows
keep an accessible dialog name
- the Artifacts view leads with the Artifacts tab and always lands on it;
Project Files and Task Documents are secondary tabs (the old auto-select
effect is gone)
- mobile tab buttons rendered at mismatched heights (two-line "Project
Files"/"Task Documents" grew past 44px while one-line "Artifacts" stayed
at 44px); tabs now pin to the uniform 44px control height with
non-wrapping labels in a scrollable row
Verified live: window dragged (264,146 -> 144,164) and resized
(1024x720 -> 872x618) in a real browser; mobile 390x844 lands on Artifacts
with uniform tab buttons.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Video was registrable but effectively unusable, and HTML/PDF deliverables
had no first-class path from agents to the gallery.
- media route now serves HTTP byte ranges (Accept-Ranges, 206 +
Content-Range, 416 on unsatisfiable) so <video>/<audio> seeking works
and Safari plays media at all
- video attachments (mp4/webm/mov, 100MB cap vs 5MB for other types)
bridge into the artifact registry like images; multer transport ceiling
raised to 100MB with per-type caps enforced in the store
- fn_artifact_register path payloads are signature-validated for video
(ftyp box / EBML header) and PDF (%PDF- prefix), mirroring images
- HTML doc artifacts (mimeType text/html) render as live sandboxed
iframe previews by default in the doc viewer, with a Preview/Source
toggle and the same FileEditor edit mode
- executor/heartbeat/planning prompts and tool descriptions now cover
the full type matrix: images, videos, audio, HTML mockups, PDFs, and
markdown docs, each with the registration recipe
Verified live: range requests (200/206/416) via curl, an ffmpeg-generated
mp4 playing to completion in the gallery lightbox, and an interactive
HTML mockup rendering in the sandboxed preview.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Agents could never get screenshots/wireframes/mocks into the Artifacts view:
fn_artifact_register was gated on assignedAgentId (never set in default
ephemeral mode), the only image payload source was inline base64, and no
prompt ever told agents to register visual deliverables.
- always expose fn_artifact_register to executor sessions ("executor" author
fallback), resolve relative paths against the task worktree, and default
taskId to the executing task (heartbeat task lane too)
- add a `path` payload source: file read with 50MB cap, extension MIME
inference, PNG/JPEG/GIF/WebP signature + SVG sniff validation, persisted
through managed artifact storage
- executor/heartbeat/planning prompts + engine-tools reference now instruct
agents to register screenshots, wireframes, mockups, and recordings
- new ArtifactsGallery: Images/Docs/PDFs/Videos/Audio/Other category sections
and filter chips, visual tile grid + lightbox, embedded PDF viewer, audio
player rows, download rows; mobile-responsive down to the 768px breakpoint
- doc artifacts open a full viewer rendered as markdown by default with an
in-place edit mode using the shared CodeMirror FileEditor; persisted via new
GET/PATCH /api/artifacts/:id + TaskStore.updateArtifact and live-refreshed
through the new artifact:updated SSE event
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Prevent the controlled FileEditor from clobbering fast local typing (including newlines) with a stale content prop echoed back from an async parent update loop.
- Track the last emitted content and a small rolling set of recently emitted self-echoes in FileEditor.
- Skip re-syncing the CodeMirror document when the incoming content prop matches a known self-echo instead of a genuinely external change.
- Clamp the previous selection into the new document length when an external content change is applied, so the caret stays valid.
- Add regression tests covering stale self-echo renders (markdown and non-markdown files) and external content changes with clamped caret preservation.
Files changed:
packages/dashboard/app/components/FileEditor.tsx | 34 +++++++++++++--
.../app/components/__tests__/FileEditor.test.tsx | 49 ++++++++++++++++++++++
2 files changed, 80 insertions(+), 3 deletions(-)
Fusion-Task-Id: FN-7808
Fusion-Task-Lineage: 08d4f826-f637-428c-8b67-dacdb878bf4a
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Reclassifies Gemini usage fetch outcomes so unconfigured/unauthenticated Gemini no longer shows a noisy error card in the usage dropdown; transient failures of a configured token still surface as errors.
- fetchGeminiUsage() in packages/dashboard/src/usage.ts now sets status to `no-auth` (instead of `error`) for unsupported auth types (api-key/vertex-ai) and for HTTP 401/403 auth-expired responses, so fetchAllProviderUsage omits Gemini from the aggregate list in those cases
- HTTP 5xx, network, timeout, and parse failures for a configured Gemini token remain `error` and visible, per the existing FN-7798 keep-auth-expired-visible convention for other providers
- Added FNXC:UsageProviders comments documenting why Gemini deliberately diverges from that convention
- Updated packages/dashboard/src/__tests__/usage.test.ts to cover the new no-auth classification
- Added changeset .changeset/fn-7806-gemini-usage.md (patch) documenting the fix for release notes
Files changed:
.changeset/fn-7806-gemini-usage.md | 7 +
packages/dashboard/src/__tests__/usage.test.ts | 211 +++++++++++++++----------
packages/dashboard/src/usage.ts | 15 +-
3 files changed, 148 insertions(+), 85 deletions(-)
Fusion-Task-Id: FN-7806
Fusion-Task-Lineage: e86b23ea-14e9-472d-8b44-3951fa02ae6c
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Mobile Settings footer now shows a compact "vX.Y.Z" version label instead of the full "Version X.Y.Z" wording, freeing horizontal space on small viewports while desktop/tablet keep the full label.
- SettingsModal renders settings.footer.versionShort ("v{{version}}") when viewportMode === "mobile", falling back to the existing settings.footer.version label otherwise
- Added settings.footer.versionShort translation key to en and fr locale files
- Updated settings-mobile tests: renamed/adjusted the mobile version test to assert the compact label and added a new test asserting the full label is kept outside the mobile viewport
- Added a patch changeset documenting the fix
Files changed:
.changeset/fn-7804-mobile-settings-version.md | 7 +++++++
packages/dashboard/app/components/SettingsModal.tsx | 10 +++++++++-
.../app/components/__tests__/settings-mobile.test.tsx | 18 +++++++++++++++---
packages/i18n/locales/en/app.json | 1 +
packages/i18n/locales/fr/app.json | 1 +
5 files changed, 33 insertions(+), 4 deletions(-)
Fusion-Task-Id: FN-7804
Fusion-Task-Lineage: 3ff5af53-3242-44b8-b8d8-6b81f416a333
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Two store-comments refinement tests located the auto-created refinement task by
`title.includes("Refinement")`, but FN-7165 (34efa8b89) intentionally changed
refinement titles to `${sourceId}: ${feedback}` for traceability. The refinement
task is still created correctly (the length assertion passes); only the
identifying substring was stale, so the tests failed on `expect(refinement)
.toBeDefined()`. Match the refinement by its feedback-derived title instead.
No product behavior change.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
updateTask now performs the explicit updates.prompt write (and its File Scope
validation) BEFORE committing the task row / task.json, instead of after. A
failed prompt write (EACCES/EISDIR/disk-full) or invalid File Scope now aborts
the whole update atomically rather than leaving the field changes committed with
a stale PROMPT.md and returning an error — the write counterpart to the
read-resilience guards in getTask/updateStep. Adds an atomicity regression test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When a task's steps live only in an unreadable PROMPT.md, updateStep still can't
resolve step 0 — but it now throws an error naming PROMPT.md as the cause instead
of a misleading "task has 0 steps". The operation genuinely cannot succeed (no
step data), so this is accurate reporting, not a silent success. The reachable
reset path (which derives its loop from getTask's steps) remains unaffected.
Adds a test asserting the PROMPT.md-attributed error.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- store.ts: guard parseStepsFromPrompt in listTasks and searchTasks too, so one
unreadable PROMPT.md can't reject the Promise.all and 500 the whole board
list/search (CodeRabbit). Matches the getTask fallback.
- update-check.ts: isHomebrewInstall now resolves symlinks and matches the real
Cellar/opt install roots, fixing Intel-macOS Homebrew detection that only
checked /usr/local/Homebrew/ (brew's repo dir) and would have shown npm/sudo
guidance instead of `brew upgrade` (CodeRabbit).
- task-detail-prompt-resilience.test.ts: extend to assert the invariant across
all surfaces — listTasks(slim)/searchTasks, reopen-to-todo moveTask
(resetPromptCheckboxes), and deleteTask — not just getTask/updateTask/archive
(CodeRabbit; Surface Enumeration rule).
- serve.test.ts: add SIGINT/SIGTERM exit-code assertions (130/143) so the serve
path's POSIX exit contract can't regress independently of daemon (CodeRabbit).
- update-check.test.ts: add Intel-Homebrew remediation test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
#5 root cause (reproduced): getTask — the shared load for the entire per-task
API — plus the mutation helpers updateTaskUnlocked, updateStep,
readPromptForArchive, and resetPromptCheckboxes all read PROMPT.md unguarded.
An unreadable PROMPT.md (root-owned from a prior `sudo` run -> EACCES, PROMPT.md
being a directory -> EISDIR, transient FS error) threw and 500'd every per-task
operation (GET/DELETE/PATCH/retry/reset/archive) for every task, while the
PROMPT.md-free board list and create kept working. These reads are now
best-effort: degrade (empty prompt / unsynced steps / skipped cosmetic sync)
and log, so a PROMPT.md hiccup can never brick task management. Added a symptom-
verification test that forces EISDIR and asserts getTask/updateTask/archiveTask
still succeed.
#10c: the dashboard badge-snapshot cache only evicted on hard-delete, so
archived tasks were re-cached via task:updated and retained for the daemon's
lifetime — a slow memory leak. New isBadgeEligibleTask predicate gates the
create/update listeners so archived tasks are evicted (matching the startup
prime's includeArchived:false). Added a unit test for the invariant.
Updates the #5 changeset to cover the real fix; adds a badge-eviction changeset.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Addresses three user-reported bugs:
- API 500 diagnosability: rethrowAsApiError now preserves the original error
as Error `cause` and the /api boundary logs stack + cause for 5xx, so the
opaque "task write API returns 500 for every task" failures are traceable
(client body stays generic in production).
- In-app "Update now": detect EACCES/EPERM install failures and return
actionable remediation (sudo fn update / reinstall without sudo / brew
upgrade) instead of raw npm stderr; do not retry --force for this class.
- Daemon restart: `fn daemon` and `fn serve` exit 128+signal (SIGTERM=143,
SIGINT=130) on signal-initiated shutdown so Restart=on-failure restarts a
memory-pressure kill. Interactive `fn dashboard` TUI intentionally unchanged.
Adds regression tests (update-check EACCES/EPERM, daemon exit codes) and three
@runfusion/fusion patch changesets.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Only show usage meters for AI providers the user has actually configured, instead of surfacing entries for providers with no meterable data.
- fetchGitHubCopilotUsage now demotes GitHub's 404 "No Copilot subscription found" response (both the Fusion-credential HTTP path and the gh-CLI fallback path) to a `no-auth` status instead of `error`, so it is treated as no meterable entitlement.
- fetchAllProviderUsage's existing `status !== "no-auth"` filter now also excludes these no-entitlement Copilot results, so they no longer appear in the usage list.
- Configured-but-failing providers (expired auth returning 401/403, transient HTTP 5xx, or other errors) keep `status: "error"` and remain visible with their diagnostic message.
- Added regression tests covering: Fusion-credential 404 omitted, Fusion-credential 500 surfaced as error, gh-CLI 404 omitted, gh-CLI 401 surfaced as "GitHub auth expired" error.
- Added a changeset documenting the usage-view behavior change as a patch/fix.
Files changed:
.changeset/fn-7798-usage-configured-providers.md | 7 +++
packages/dashboard/src/__tests__/usage.test.ts | 71 ++++++++++++++++++++++--
packages/dashboard/src/usage.ts | 19 ++++++-
3 files changed, 90 insertions(+), 7 deletions(-)
Fusion-Task-Id: FN-7798
Fusion-Task-Lineage: 18835b48-68f3-48aa-a03c-cc85772778a9
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Replace the boolean isGitRepository() check with a tri-state Git detection so environmental git failures (dubious ownership, missing git binary, timeouts) are no longer misreported as "not a Git repository", which previously blocked all task execution in valid repos and survived engine restarts.
- Add detectGitRepository() in worktree-pool.ts returning repo / not-repo / error (with reason: dubious-ownership, git-missing, timeout, unknown), classified from git's stderr; bound the git rev-parse call with a 10s timeout and maxBuffer; keep isGitRepository() as a backward-compatible wrapper
- Route the executor dispatch preflight guard through detectGitRepository(): only emit the original "not a Git repository / run git init" fatal on a positive not-repo verdict; on error, throw a distinct accurate error naming the real git failure, including the safe.directory remedy for dubious ownership
- Route the in-process runtime startup warning through the same tri-state detection so it only warns "not a Git repository" on a positive not-repo verdict
- Add a regression test locking extractWorktreeConflictInfo() to NOT misclassify a dubious-ownership git worktree add failure as not-git-repo
- Add targeted tests across worktree-pool, executor-worktree, and in-process-runtime test suites covering repo/not-repo/dubious-ownership/git-missing/timeout classifications on Windows OneDrive-style and POSIX paths
- Add changeset and a docs/solutions/logic-errors write-up of the false-negative root cause and fix
Files changed:
.changeset/fn-7799-git-detection-false-negative.md | 7 +++
.../logic-errors/git-detection-false-not-repo.md | 54 ++++++++++++++++
.../engine/src/__tests__/executor-worktree.test.ts | 61 +++++++++++++++++++
.../engine/src/__tests__/worktree-pool.test.ts | 71 +++++++++++++++++++---
packages/engine/src/executor.ts | 38 +++++++++---
.../runtimes/__tests__/in-process-runtime.test.ts | 53 ++++++++++++++--
packages/engine/src/runtimes/in-process-runtime.ts | 16 ++++-
packages/engine/src/worktree-pool.ts | 66 ++++++++++++++++++--
8 files changed, 334 insertions(+), 32 deletions(-)
Fusion-Task-Id: FN-7799
Fusion-Task-Lineage: 25a84283-bf47-472b-8a98-a10bf7e494de
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Fix ProviderIcon so Grok model-shaped provider strings (e.g. grok-4.5, grok-cli/grok-4-fast) render the xAI brand icon instead of falling back to the generic Cpu icon.
- ProviderIcon now falls back to inferProviderIconKey(provider) when an exact providerConfig lookup misses, before defaulting to the Cpu icon
- Added a changeset (patch) documenting the fix for dashboard provider icon surfaces
- Added regression tests covering exact xAI/Grok keys, inferred Grok model-id strings, non-Grok inferred providers (e.g. gpt-5.5), and confirming unknown/empty providers still fall back to Cpu without the xAI icon
Files changed:
.changeset/fn-7801-grok-xai-logo.md | 7 ++++++
packages/dashboard/app/components/ProviderIcon.tsx | 11 ++++++++-
.../app/components/__tests__/ProviderIcon.test.tsx | 27 ++++++++++++++++++++++
3 files changed, 44 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-7801
Fusion-Task-Lineage: e81f15f6-d295-4f6d-b21f-7cf58854457f
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Closes the durability gap behind "grok chat returns empty replies". In a source
checkout the dashboard resolved the staged CLI tsup bundle
(packages/cli/dist/plugins/<id>/bundled.js), which resolvePluginEntryPath prefers
verbatim with no freshness check. The FN-7779 dev prebuild rebuilds each plugin's
own plugins/<id>/dist but never the staged tsup bundle, so a source-only plugin
fix ran stale until a manual `pnpm build`.
getCandidatePluginDirs now probes the live workspace source dir (<repo>/plugins/<id>)
before the staged bundle, so dev loads the freshness-checked live plugin (dist-vs-src),
self-healing even when the prebuild is skipped. The global-staged dir stays first,
so published installs (no workspace dir) are unaffected — asserted by the retained
global-install regression test plus a new source-checkout preference test.
Verified on a live dashboard: grok chat streams text and the loader now writes its
reload copies under plugins/fusion-plugin-grok-runtime/dist, not the staged bundle.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two failure modes wedged the PR-based merge flow (dormant under the
current direct-merge config, but the rollback target):
1. Empty-diff branches ("No commits between ...") were marked `failed`,
pinning the serial merge slot + file leases on a task that is a
legitimate no-op. Now finalized as a terminal DONE via
`finalizeNoOpMergeTask`, mirroring the engine's canonical
`noOpResult` decision (merger-ai.ts).
2. A stale-base PR that GitHub reports CONFLICTING never becomes
mergeable on its own (nothing in the PR path rebases the head), so
`awaiting-pr-checks` waited unbounded — no escape, because the PR
path never incremented `mergeRetries`. Now each conflicting poll
counts against `mergeRetries`, so the existing
`getInReviewStallReason` "merge-retries-exhausted" escape disposes
the task after `maxAutoMergeRetries` cycles. Pending/behind PRs
still wait (checks legitimately run; "behind" is fixed by rebase).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Greptile/CodeRabbit review of the chat slash-command framework:
- Composer-wipe race: ChatView and TaskPlannerChatTab cleared the composer
inside the command's success callback, silently wiping any text the user
typed while the command was in flight. Clear on submit (before the network
round-trip) instead — consistent with normal chat send, which also clears
immediately and does not restore on failure.
- Attachments were silently dropped when dispatching a slash command in
ChatView (clearing the composer revokes staged attachment URLs). Block
dispatch with a warning toast when attachments are staged.
- CHAT_COMMANDS is now a readonly array; helper signatures accept
readonly ChatCommand[].
- The planner command menu (commands-only) used skill-menu aria-label/empty
copy; use command-specific copy instead.
Add regression tests: in-flight text survives command success, composer
clears on submit even on failure, attachment dispatch is blocked, and the
planner command menu uses command-specific accessible copy.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a chat command registry and the /steer command for task-bound
chats. Commands are triggered by '/' and dispatch through a common
match/dispatch path alongside existing skills autocomplete.
Operator follow-up on the centered 960px column: the Memory panes should
stretch to the full width of the view; the editor benefits from the space.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Working/Insights/Engines tabs now render a centered 960px content
column instead of leaving the right half of wide viewports empty.
- The file editor's collapsed chevron-only toolbar (an unlabeled mystery
control) now always shows its labeled actions in MemoryView.
- 'Process dreams' / 'Auto-summarize memory' rows share one aligned
layout, sentence case, and a tooltip explaining dream processing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The reliability GET/reset handlers referenced getScopedStore, which is
defined inside setupBadgeWebSocket and is not visible in the createServer
scope where these handlers live — so the scoping change did not typecheck.
Switch to the in-scope resolveProjectScopedStore helper (used by the other
realtime endpoints), which also routes through engineManager for correct
per-project resolution.
Guard store resolution with try/catch returning a targeted 500, mirroring
the project SSE handler, instead of falling through to the generic error
handler. Add project-scoping regression tests: GET reads the project store,
GET without projectId falls back to root, and reset writes the project store.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reliability GET and reset endpoints were always using the server's
root store, ignoring projectId. Mirror the Command Center pattern by
using getProjectIdFromRequest and getScopedStore so multi-project
servers report per-project reliability stats.
Refs FUX-042
The panel carries the shared .card class, and '.card { position: relative }'
loads after the popover stylesheet, so equal specificity let bundle order
strip the panel's position: fixed — it then flowed inside the documents
viewer and rendered clipped at the bottom-right of the viewport, far from
the selection. A .selection-comment-panel.card rule restores fixed
positioning immune to order; the panel's left is now a width-aware clamp
(no half-offscreen composer near viewport edges), top is clamped near the
bottom, and the textarea focuses with preventScroll so opening the
composer no longer scrolls the selected content out of view.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The global .btn:active press-feedback transform (scale 0.97) overrode
the selection-comment trigger's positioning translate, teleporting the
fixed-position button mid-press so the click never landed and the
composer silently never opened. The :active state now restates the
translate alongside the scale on desktop and mobile offsets. Document
previews also gain a subtle Read-only badge explaining select-to-comment.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Operator design follow-up: even grouped, nine outlined chips read noisy.
Option controls and the secondary attach/Fast pair are now borderless
ghost buttons in muted text that gain a subtle fill and border on
hover/focus/open, leaving Save as the row's single emphasized control.
Scoped to the quick-entry action row; shared .btn/.dep-trigger rules
untouched.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Quick-add composer: single divider, option chips grouped left
(workflow, priority, steps, deps, models, node, agent, GitHub) and
primary actions right-aligned (attach, fast, Save last).
- Task cards get a visible kebab button (hover on desktop, always on
mobile) opening the same context menu as right-click, which was
previously undiscoverable.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Live verification showed .setup-wizard-isolation-option (0,1,0) losing
the cascade to the global .form-group label rule (0,1,1), so the cards
still rendered ALL-CAPS. Add a higher-specificity neutralization rule.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- AI Setup: connected providers now render first; 'Skip for now' hides
once a provider is connected; footer buttons share one size.
- GitHub step: state-driven (no install/pitch content when connected),
one type scale, real <code> literals, per-OS gh instructions behind a
disclosure, single skip affordance; gh/OAuth status revalidates on
window focus and OAuth relogin so later steps stop showing stale
'not connected'.
- Project step: onboarding-driven opens keep 'Step 3 of 5' context;
runtime-node picker no longer lists 'Local node' and 'local (local)'
duplicates and hides itself when only the local node exists;
isolation-mode cards no longer render a stretched native radio or
uppercased descriptions.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The GitLab Configuration disclosure was the only .settings-content
child without the form-group horizontal gutter, so its bordered box and
right-flushed checkbox leaked to the pane edge. The footer action rail
was a no-wrap flex with zero slack, clipping Save whenever extra footer
content appeared; it now wraps on desktop (mobile scroll rail intact).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The connection status pill rendered the host kind twice ('Desktop
Desktop local mode') with mismatched font sizes. The kind is now folded
once into the summary ('Desktop · Local mode') across all pill states.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Grok CLI chat failed two different ways depending on the surface:
1. Default (no-project) chat errored with "requires the bundled Grok CLI
runtime". The default ChatManager was handed a bare PluginLoader, but Grok
routing (deriveGrokRuntimeHintForNoVisibleKey -> resolveRuntime) needs a
PluginRunner's getRuntimeById/createRuntimeContext; the unguarded call threw
"getRuntimeById is not a function". New resolveChatManagerPluginRunner()
prefers the engine's PluginRunner (same runner the project-scoped path uses),
falling back to the loader only in UI-only mode.
2. Project-scoped chat returned empty replies. The CLI-bundled Grok plugin
(packages/cli/dist/plugins/.../bundled.js, gitignored) was stale vs the
FN-7796 single-JSON adapter source; the running server loads that bundle,
not the plugin's own dist. `pnpm build` regenerates it. Noted in the
changeset that the freshness guard only warns and the dev prebuild does not
rebuild the CLI tsup bundle.
Verified end-to-end on a live dashboard: both default and project-scoped
grok-cli/grok-4.5 chats now stream thinking + text.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Grok stale-dist bug was possible because the dev/build path never
refreshed plugin dist:
- The `client` prebuild (default `pnpm dev dashboard`) rebuilt only
@fusion/core + @fusion/engine + @fusion/dashboard, never plugins.
- The FN-6638 stale-dist startup warning only scanned packages/, never
plugins/, so a source-ahead plugin dist ran phantom-old with no warning.
Changes:
- build-workspace.mjs: add `--plugins-only` to plan/build just the plugins
that changed, reusing the existing content-hash skip cache (cheap no-op when
unchanged).
- scripts/dev-prebuild-client.mjs: new orchestrator — fast core/engine/
dashboard build, then incremental changed-plugin rebuild. The `client`
prebuild now runs this single cross-platform command.
- dist-freshness.mjs: scan plugin roots (plugins/, plugins/examples/) so a
stale plugin dist is warned like a stale package dist; the warning names the
plugin dir.
Verified: --plugins-only plans only plugins, skips unchanged on the second
run, and re-plans exactly the one plugin whose source changed. All script and
CLI lib tests pass.
Fusion-Task-Id: FN-7779
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds a per-lane thinking-level selector to every fallback model picker across global, project, and workflow Settings surfaces, wiring them into the existing null-as-delete save paths.
- Add inline thinking-level dropdowns to the global Fallback Model, workflow-declared planning/validator fallback lanes, and the project-scoped Title Summarizer fallback picker via CustomModelDropdown's `showThinkingLevel`.
- Route `planningFallbackThinkingLevel`/`validatorFallbackThinkingLevel` through workflow settings PATCH and `titleSummarizerFallbackThinkingLevel` through project settings save-split, both with null-as-delete semantics on reset.
- Move Title Summarizer fallback out of the workflow-declared model pairs into a dedicated project-scoped lane in ProjectModelsSection so its thinking companion isn't tied to workflow settings.
- Update WorkflowSettingsPanel to surface the new fallback thinking companion keys.
- Extend/adjust tests (settings-save-split, settings-sections, SettingsModal.models-auth, WorkflowSettingsPanel, core settings-migration) to cover the new selectors and save routing.
- Document the fallback thinking-level runtime behavior in docs/settings-reference.md and docs/dashboard-guide.md.
- Add a minor changeset describing the new fallback thinking-level selectors.
Files changed:
.changeset/fn-7795-fallback-thinking-selectors.md | 7 +
docs/dashboard-guide.md | 4 +-
docs/settings-reference.md | 10 +-
packages/core/src/__tests__/settings-migration.test.ts | 14 +-
packages/dashboard/app/__tests__/settings-save-split.test.ts | 35 +++++
packages/dashboard/app/__tests__/settings-sections.test.tsx | 163 ++++++++++++++++++++-
packages/dashboard/app/components/WorkflowSettingsPanel.tsx | 6 +-
packages/dashboard/app/components/__tests__/SettingsModal.models-auth.test.tsx | 11 +-
packages/dashboard/app/components/__tests__/WorkflowSettingsPanel.test.tsx | 38 +++++
packages/dashboard/app/components/settings/save-split.ts | 9 +-
packages/dashboard/app/components/settings/sections/GlobalModelsSection.tsx | 8 +-
packages/dashboard/app/components/settings/sections/ProjectModelsSection.tsx | 58 ++++++--
12 files changed, 328 insertions(+), 35 deletions(-)
Fusion-Task-Id: FN-7795
Fusion-Task-Lineage: ec990d47-defe-4717-993a-56988afe8d7d
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
Adds fallbackThinkingLevel plumbing so, when Fusion swaps from a primary model to a configured fallback model (executor, validator/reviewer, merger, planning, title-summarizer, heartbeat, and workflow-step lanes), the fallback's own configured thinking level is applied instead of silently reusing the primary lane's level.
- Add fallbackThinkingLevel option to AgentRuntimeOptions (agent-runtime.ts), AgentOptions (pi.ts), and ReviewOptions (reviewer.ts)
- Add per-lane resolvers: resolveExecutorFallbackThinkingLevel, resolvePlanningFallbackThinkingLevel, resolveValidatorFallbackThinkingLevel, resolveTitleSummarizerFallbackThinkingLevel, resolveMergerFallbackThinkingLevel (agent-session-helpers.ts), each following fallback-provider precedence and falling back to the primary lane/default thinking level when unset
- Export new resolvers from packages/engine/src/index.ts
- Apply the resolved fallback thinking level in createFnAgent's applyThinkingLevelIfSupported once a session has swapped to the fallback model (pi.ts)
- Wire fallbackThinkingLevel through executor session creation (workflow-step, task validator, child-agent, and main executor session paths), merger session creation, and heartbeat session creation
- Promote the fallback thinking level alongside the fallback model/provider when the no-visible-key Grok CLI fallback is promoted to primary, so the cleared fallback pair doesn't leave the session on the superseded primary's thinking level
- Route workflow-step fallback thinking level by which fallback candidate (validatorFallback vs globalFallback) actually matched
- Document fallbackThinkingLevel runtime-swap behavior in docs/settings-reference.md
- Add minor changeset for @runfusion/fusion
- Add regression tests covering fallback thinking-level resolution and application (agent-session-helpers.test.ts, pi.test.ts) and a shared test helper (executor-test-helpers.ts)
Files changed:
.changeset/fn-7794-fallback-thinking-level.md | 7 ++
docs/settings-reference.md | 3 +
.../src/__tests__/agent-session-helpers.test.ts | 38 ++++++
.../engine/src/__tests__/executor-test-helpers.ts | 23 ++++
packages/engine/src/__tests__/pi.test.ts | 136 +++++++++++++++++++++
packages/engine/src/agent-heartbeat.ts | 3 +-
packages/engine/src/agent-runtime.ts | 5 +
packages/engine/src/agent-session-helpers.ts | 54 ++++++++
packages/engine/src/executor.ts | 31 ++++-
packages/engine/src/index.ts | 5 +
packages/engine/src/merger.ts | 7 +-
packages/engine/src/pi.ts | 16 ++-
packages/engine/src/reviewer.ts | 6 +
13 files changed, 327 insertions(+), 7 deletions(-)
Fusion-Task-Id: FN-7794
Fusion-Task-Lineage: c94d621a-ccbd-42b2-9fe6-cb619418ad90
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>
