Triage runs through pi-claude-cli still hang silently for 20+ minutes after
only 2 thinking-deltas before the engine's StuckTaskDetector kills them. We
can't tell from agentLogEntries whether the subprocess crashed, the MCP
server failed to attach, or Claude is in extended thinking that doesn't
stream.
Adds stderr lines tagged [pi-claude-cli] for:
- spawn (pid, model, mode, prompt sizes, mcp on/off)
- init (session id, model, permissionMode, mcp_servers status list)
- first-stdout-line (latency from spawn)
- tool_use (per top-level tool, with piKnown flag)
- break-early (when we kill at message_stop)
- close (exit code, signal, elapsed, broken-flag)
No behavioural change. Restart the dashboard, trigger triage, and these
lines tell us which phase silence sets in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Delete obsolete AppModals code that was accidentally left in ModelOnboardingModal.tsx
- Restore the Task type import used by the onboarding modal after dead code cleanup
- Drop the unused SettingsModal.css import tied to removed code paths
- Extend ModelOnboardingModal tests to guard against dead export regressions
- Track plain-text render mode per assistant message (including streaming output) in ChatView state
- Add inline mobile-only Eye/EyeOff toggle buttons on assistant bubbles to switch markdown vs plain text per message
- Hide the header markdown/plain toggle on mobile and polish inline toggle sizing, focus, and hover styles in ChatView.css
- Extend ChatView tests with mobile viewport coverage for toggle rendering, per-message isolation, round-trip toggling, and mobile CSS contract checks
- Add generated engine-tools reference and refresh extension/capabilities docs from source definitions
- Enhance sync-fusion-skill-tools script to verify and maintain skill documentation consistency
- Expand skill-sync tests and enforce sync:fusion-skill:check in the workspace test pipeline
- Add FN-2569 changeset documenting the published @runfusion/fusion patch update
- Replace app/api.ts with a compatibility barrel that re-exports the legacy client API surface from app/api/legacy.ts
- Move the existing dashboard API implementation into app/api/legacy.ts and update static-analysis tests to include the new module path
- Update proxy wildcard routing to use /proxy/:nodeId/{*splat}, keeping explicit proxy handlers ahead of the fallback route
- Reject proxying to local nodes with a consistent 400 response and expand proxy route tests plus route-ordering compatibility notes
- Set the combobox search wrapper background to var(--surface) to prevent option list bleed-through behind the input.
- Update the dropdown header background from var(--bg) to var(--surface) for a consistent opaque top section.
- Add a CSS regression test in CustomModelDropdown.test.tsx that asserts the search wrapper rule includes an opaque surface background.
- Map the pi-claude-cli provider key to the Claude CLI icon config in ProviderIcon.
- Reuse Anthropic tokenized provider color and accessible label for the new alias.
- Add dashboard tests asserting icon rendering, color/label metadata, and mixed-case alias normalization.
- Enrich TaskTokenStatsPanel with timing extraction, workflow runtime summaries, and longest-event highlights from task logs and workflow results
- Add an execution details section for mode, runtime status, step progress, retries, recovery state, and runtime links
- Wire TaskDetailModal to pass the working task into the stats panel so the Stats tab renders execution context with token usage
- Extend TaskDetailModal and TaskTokenStatsPanel tests (plus panel CSS) to cover the new stats sections and loading/empty/token states
- Reset memory backend status to unresolved during fetches so stale capabilities are never treated as authoritative
- Gate MemoryView read-only and qmd warning states on resolved backend status and add an explicit checking state for qmd availability
- Update SettingsModal memory editor gating so writes remain enabled while backend status is still loading, then enforce resolved read-only state
- Expand dashboard tests for useMemoryBackendStatus, MemoryView, and SettingsModal to lock in loading, warning, and refresh behavior
The engine already runs a StuckTaskDetector with a configurable per-task
heartbeat (default 1 hour) and aborts triage/executor sessions through
AbortSignal when it judges them quiet. pi-claude-cli forwards that signal to
the subprocess. The 300s subprocess-level inactivity timeout was a redundant,
much-tighter shadow that kept killing Sonnet 4.6 mid-thinking on the 40k-char
triage prompt — exactly what the engine-level detector is designed not to do.
Move the local timeout to 30 minutes purely as a last-resort guard for embeds
that don't pass an abort signal. Stuck-detection responsibility now lives in
the caller, where it can be configured per task.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Prevent TaskCard footer row metadata from wrapping by enforcing nowrap layout
- Make files-changed chip flex safely and truncate long text with ellipsis while preserving icon visibility
- Keep timer chip placement anchored to the right side of the footer metadata row
- Add TaskCard and board-mobile regression tests asserting footer child order and CSS layout guarantees
Sonnet 4.6 with extended thinking on the triage prompt (~40k chars + AGENTS.md
+ skills) can take >3 minutes between thinking deltas. The 180s inactivity
timeout was killing those subprocesses before they could finish PROMPT.md +
fn_review_spec, leaving triage looping into the fallback model. Bump to 5
minutes.
Test mock timings updated to match.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Add a new renderBelowModelConfiguration prop in TaskForm and render it beneath the Model Configuration section inside More options
- Update TaskDetailModal to use the new slot so the Source Issue edit block appears below model settings instead of near the primary section
- Adjust TaskDetailModal source-issue edit tests to expand More options before interacting with source fields
- Add ordering assertions in TaskForm and TaskDetailModal tests to verify the injected block sits between Model Configuration and Workflow Steps
Root cause of "fn_review_spec was never called" tracked through the diagnostic
chain: the prompt actually sent to triage agents has zero `fn_*` tokens —
because resolveAgentPrompt("triage", ...) returns the BUILTIN_AGENT_PROMPTS
default-triage template (TRIAGE_PROMPT_TEXT in core/agent-prompts.ts), and that
template was forked from an older version that never had the "MUST call
fn_review_spec()" workflow nor any fn_-prefixed tool names. The fallback
`|| TRIAGE_SYSTEM_PROMPT` in engine/triage.ts never fires because the core
template is non-empty.
So the model writes PROMPT.md, doesn't see any instruction to review it, and
ends. zai/glm-5.1 happened to call fn_review_spec from training-pattern
inertia; Sonnet via pi-claude-cli stopped at write — same prompt, same bug.
Replace TRIAGE_PROMPT_TEXT with the engine's up-to-date TRIAGE_SYSTEM_PROMPT
verbatim (fn_-prefixed tools, fn_review_spec workflow, subtask breakdown,
project-commands handling, frontend UX criteria injection). Also remove the
diagnostic-only console.error lines added during this debugging session — the
core fix is now elsewhere and the noise isn't worth keeping.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
User reports "no custom tool refs to rewrite (tools=18)" — meaning Context.tools
is populated but the prompt body doesn't contain any of the names. Need the
first 200 chars of the prompt and a sample of custom-tool names to figure out
whether the engine's customPrompt is reaching us at all or whether names
differ from what we expect.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Preserve overdue nextRunAt when schedule updates only touch non-cadence fields
- Recompute nextRunAt only when cadence changes, schedules are re-enabled, or nextRunAt is missing
- Sync memory dreams automation during ProjectEngine startup before CronRunner begins ticking
- Add core/engine regression coverage and a patch changeset for @runfusion/fusion release notes
Adds stderr lines so we can verify the rewrite is actually firing on a given
session ("rewrote N custom tool ref(s) [fn_review_spec×3, ...]") vs. silently
no-opping (no tools, no matches). Helps distinguish "fix not deployed yet"
from "fix not effective" without redeploying instrumentation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Split timer calculation paths so in-progress cards show live elapsed time since entering in-progress
- Make done cards show fixed processing duration from start to completion instead of growing post-completion elapsed time
- Improve timer tooltip and aria-label copy for clearer in-progress and done semantics
- Expand TaskCard tests to cover boundary formatting, fixed done-duration behavior, and stable timer output as time advances
Triage system prompts read "MUST call fn_review_spec()" but Sonnet 4.6 routed
through pi-claude-cli writes PROMPT.md and silently skips the call — even with
the addendum explaining the deferred-tool protocol. Confirmed by the FN-2564
agent log: model called other MCP tools (fn_memory_search, fn_task_list) fine
but consistently never reached fn_review_spec, leaving triage looping on
"fn_review_spec was never called" and falling back to zai/glm-5.1 every time.
Rewrite bare `fn_*` (and any non-built-in custom tool name) references in the
system prompt to their `mcp__custom-tools__fn_*` form before sending. The
prompt now literally says "call mcp__custom-tools__fn_review_spec()" so the
model has no inference step, and the deferred-tool reminder Claude Code injects
matches verbatim. Word-boundary safe (won't touch fn_review_specifier) and
idempotent (won't double-prefix already-MCP-named occurrences).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Triage system prompts read "MUST call fn_review_spec()" but Sonnet 4.6 routed
through pi-claude-cli writes PROMPT.md and silently skips the call — even with
the addendum explaining the deferred-tool protocol. Confirmed by the FN-2564
agent log: model called other MCP tools (fn_memory_search, fn_task_list) fine
but consistently never reached fn_review_spec, leaving triage looping on
"fn_review_spec was never called" and falling back to zai/glm-5.1 every time.
Rewrite bare `fn_*` (and any non-built-in custom tool name) references in the
system prompt to their `mcp__custom-tools__fn_*` form before sending. The
prompt now literally says "call mcp__custom-tools__fn_review_spec()" so the
model has no inference step, and the deferred-tool reminder Claude Code injects
matches verbatim. Word-boundary safe (won't touch fn_review_specifier) and
idempotent (won't double-prefix already-MCP-named occurrences).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
pi-coding-agent 0.70's createCodingTools preset only includes read/bash/edit/write,
but pi-claude-cli's tool-mapping translates Claude's Glob→find and Grep→grep.
Triage sessions running through Claude CLI hit "Tool find not found" the moment
the model called Glob and looped on the error.
Compose the built-in set explicitly so every tool referenced by tool-mapping.ts
is registered. Read-only sessions also gain ls (was silently dropped earlier
when readonly was redefined upstream).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
pi-coding-agent 0.70's createCodingTools preset only includes read/bash/edit/write,
but pi-claude-cli's tool-mapping translates Claude's Glob→find and Grep→grep.
Triage sessions running through Claude CLI hit "Tool find not found" the moment
the model called Glob and looped on the error.
Compose the built-in set explicitly so every tool referenced by tool-mapping.ts
is registered. Read-only sessions also gain ls (was silently dropped earlier
when readonly was redefined upstream).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Treat devServer as a legacy alias of canonical devServerView in Experimental Features
- Canonicalize feature keys when rendering and saving settings to prevent duplicate Dev Server rows
- Preserve enabled-state compatibility by reading either key in useAppSettings
- Add regression tests for single-row rendering, alias normalization on save, and legacy hook behavior
Claude Code 2.x defers MCP tool schemas behind ToolSearch — tools are listed
but calling them by short name (e.g. fn_review_spec) returns "tool not found"
until ToolSearch loads the schema. Triage system prompts that say "call
fn_review_spec()" leave the model guessing whether to use the bare name or
the mcp__custom-tools__ form.
Append a system-prompt section listing every custom tool with its short and
MCP-prefixed names plus the ToolSearch + call protocol. Drives off the
Context.tools we now already thread through, so plain chats with no custom
tools get no addendum.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Add explicit executionMode props to TaskForm edit controls and render a Standard/Fast selector in More Options
- Track editExecutionMode state in TaskDetailModal and reset it consistently when opening, canceling, or syncing task changes
- Include executionMode in update payload diffing so standard→fast sends "fast" and fast→standard clears to null
- Add regression tests for TaskForm selector behavior and TaskDetailModal update payload handling
- Add overlap-ignore path validation and typed settings support in core schema
- Apply overlap ignore paths in scheduler overlap detection with dedicated engine tests
- Add Settings modal UI and routes handling for overlap ignore paths including path-picker feedback fixes
- Document overlap ignore paths in storage/settings docs and include a changeset for @runfusion/fusion
- Add a workflow-failed dot modifier class for failed workflow checks in TaskCard
- Apply ws-warning styling to workflow-origin failed steps while keeping regular failed steps unchanged
- Update TaskCard rendering logic to append the workflow-failed class only for workflow failed items
- Expand TaskCard tests to verify class assignment for regular failed, workflow failed, done, and pending dots
- Consolidate mobile mailbox header-actions rules for modal and view layouts with tokenized spacing
- Enforce 36px minimum height/width for action buttons and icon controls on mobile
- Normalize modal close control mobile padding and radius for consistent alignment
- Add CSS contract assertions in MailboxModal and MailboxView tests for spacing and control dimensions
When users have an external pi-claude-cli (e.g. a global `npm install -g
pi-claude-cli`, or `npm:pi-claude-cli` in ~/.pi/agent/settings.json packages),
pi's extension discovery loaded the upstream copy and shadowed our fork. The
upstream has a once-and-lock MCP-config bug that throws "Extension runtime not
initialized" during early streamSimple calls and never recovers.
Adds reconcileClaudeCliPaths in @fusion/core, used by both the daemon's
extension assembly and the engine's per-session registerExtensionProviders, to
drop any path with a `pi-claude-cli` segment that isn't our vendored fork and
prepend the vendored path. Engine resolves the fork via require.resolve and
gracefully no-ops when it isn't reachable (e.g. embedded standalone usage).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Add dismissal tombstone tracking in useBackgroundSessions to ignore stale refresh, sync, and SSE updates for dismissed session IDs
- Treat terminal session states consistently and broadcast completion on dismiss so local/session-sync state remains converged
- Keep lock-conflicted planning cancellations visible instead of falsely dismissing the session when another tab owns the lock
- Expand useBackgroundSessions regression tests to cover stale sync/SSE resurrection, refresh behavior, lock-conflict handling, and newer-authoritative restore paths
- Extend column action menus beyond Todo to include In Progress and In Review
- Add Stop All action that pauses only non-paused tasks with confirmation and success/error toasts
- Add Move All to Todo action with confirmation and partial-failure handling for bulk moves
- Thread pauseTask through useTasks, App, and Board so column menus can trigger task pausing
- Expand Column tests to cover new menu actions, disabled states, and bulk operation behavior
The previous fix regenerated the MCP config per call but still pulled the tool
set from pi.getAllTools(), which depends on the runtime binding being scoped to
the right AgentSession. Engine-spawned triage/executor sessions register custom
tools (fn_review_spec, fn_review_step) and pi-ai already passes those to the
provider via Context.tools — that's the authoritative per-session list and
bypasses any binding ambiguity. Falls back to pi.getAllTools() when the caller
doesn't populate Context.tools.
Also logs the tool names on refresh so missing tools are diagnosable from stderr.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Annotate the intentional NUL-byte regex in log-sink, drop two stale
eslint-disable directives whose underlying any-uses are gone, and swap
@ts-ignore for @ts-expect-error in routes.test.ts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
TUI quit ('q'/Ctrl+C) bypassed signal handlers via process.exit(0), and
neither shutdown path closed the HTTP server, so server.close()'s
stopAllDevServers() listener never ran. In-flight agent bash commands
(spawned detached for their own pgroup) were also never aborted, so
their subprocess trees — including vitest workers — survived as orphans.
Route the TUI quit through SIGINT so the registered shutdown handler
runs, await stopAllDevServers() in both shutdown paths, and abort
in-flight bash on every active agent session at the start of the
runtime drain so killProcessTree reaches every grandchild.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The MCP config was generated lazily once and locked, so engine session-scoped
tools (fn_review_spec, fn_review_step) never reached the Claude CLI subprocess
and triage/executor sessions failed with "unknown tool" errors. Now the config
is hashed per call and rewritten when the tool set changes.
Also adds scripts/sync-fusion-skill-tools.mjs to regenerate the SKILL.md
tool-categories block from extension.ts at build time, with a --check mode
wired into skill-sync tests so drift fails CI.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Add typed health endpoint helper in dashboard API for retrieving status metadata
- Fetch dashboard health in SettingsModal and render a non-blocking version label in the modal header
- Add settings modal heading/version styles to support the new header layout
- Expand SettingsModal tests to cover successful version rendering and graceful failure behavior
- Add mobile settings test coverage for displaying the version label
- Move mobile .toast-container positioning from bottom offset to top offset beneath the fixed header
- Tokenize horizontal toast spacing on mobile using var(--space-sm) instead of hardcoded pixel values
- Update utility mobile CSS regression test expectations for top anchoring and auto bottom behavior
- Refactor TaskCard to compute files-changed metadata once and render it through a shared footer slot
- Render elapsed time chip in the same footer row as file-change metadata when either element is present
- Update TaskCard styles to add a reusable .card-footer-row layout and align the timer chip to the row end
- Add regression coverage asserting files-changed and timer chips coexist in one footer container
- Derive task elapsed time from columnMovedAt with updatedAt/createdAt fallbacks and guard against invalid or future timestamps
- Render a clock-based timer chip on in-progress and done cards with accessible labeling and tooltip metadata
- Add TaskCard styles for the timer row/chip using design tokens, including mobile-size adjustments
- Expand TaskCard tests to cover visibility by column, invalid timestamp suppression, boundary label formatting, and 30s live refresh cadence
- Update skills adapter fallback logic to handle empty and short catalog queries without surfacing 400 failures
- Handle ApiRequestError paths in SkillsView so invalid-query responses degrade gracefully in the UI
- Add regression coverage for adapter, routes, and SkillsView behavior across empty/short query cases
- Document short-query catalog behavior in the dashboard guide and include a patch changeset for @runfusion/fusion
