- Remove ToolSearch prerequisite so custom MCP tools can be used during triage and plan generation flows
- Align built-in tool sets between provider wiring and prompt builder handling, including custom ls behavior
- Expand pi-claude-cli tests for event bridge, MCP config, prompt builder, and tool mapping regressions
- Add FN-2703 changeset and delivery documentation for the published @runfusion/fusion package
Manual version bump to 0.5.0 (changeset version produced 1.0.0 from a
single minor changeset against 0.4.1; that release was rolled back and
the 1.0.0 npm version deprecated).
Aggregates: status terminology refresh (planning/replan), Reviewer
rename, in-review pause behavior, dashboard-tui resize hardening,
dev-server experimental toggle fix, version reporting fix.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- test(FN-2618): harden process-manager fs/tmpdir mocks
- fix(cli): mark react-devtools-core external in bun compile
- fix(pty): switch to @homebridge/node-pty-prebuilt-multiarch fork
- Update TaskCard done-duration calculations to use createdAt as the start baseline for completed tasks
- Refactor TaskCard timer tests to assert against TaskStore-backed timestamps and avoid brittle elapsed-time assumptions
- Cover done-state timer behavior with expanded test scenarios for start/end timestamp combinations
- Remove unused provider code in pi-claude-cli to satisfy lint and keep the full test/build gates green
Version bump via changesets (consumed 11 changesets). Forced 0.3.0
instead of changesets' default escalation to 1.0.0 for the pre-1.0
minor bump.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Triage with claude-sonnet-4-6 via pi-claude-cli kept looping on
fn_review_spec calls that were rejected by pi's validator with
"root: must be object". Parameterless MCP tools (schema
{type:"object", properties:{}}) emit zero input_json_delta events,
so partialJson stayed "" and the catch fell through to
finalArgs = "" — a string, which TypeBox's Type.Object({}) rightly
refuses. Default empty partialJson to {} so the call lands.
Also:
- Add a 2-step reminder loop in triage before swapping to the
fallback planning model — primary models that wrote PROMPT.md
but forgot fn_review_spec recover from a nudge, no need to pay
the cold-start tax of a new triage on a different model.
- Inject @runfusion/fusion's own pi extension into dashboard/
daemon/serve sessions and propagate the path to createFnAgent
via setHostExtensionPaths so fn_* tools register globally
without requiring `pi install npm:@runfusion/fusion`.
- Drop the "historical" qualifier from replayed tool labels —
Claude was reading "TOOL RESULT (historical Read):" as
"previous session, ignore" and looping on verification.
- Remove subprocess-lifecycle stderr debug logs that landed for
hang diagnosis — root cause is fixed, the noise can go.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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>
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>
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>
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>
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>
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>
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>
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>
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>
TUI: vitest memory-guard threshold and on/off toggle now persist to
global settings (vitestAutoKillEnabled / vitestKillThresholdPct), so
they survive dashboard restarts. Stats panel shows the system-memory
used percentage next to used/free. Utilities panel exposes [+/-] to
adjust the threshold in 5% steps (50–99%).
Release: scripts/release.mjs auto-syncs a root CHANGELOG.md aggregated
from every packages/*/CHANGELOG.md, grouped by version with one
sub-block per package.
Versioning: all private @fusion/* packages joined the changesets fixed
group with the public cli + cli-alias and were aligned to 0.2.5, so
every release bumps every package and produces per-package CHANGELOG
entries that the aggregator picks up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Move all co-located *.test.* files into sibling __tests__/ directories so the
layout is consistent across packages (159 renames + content-rewrite moves).
Updates relative imports, vi.mock specifiers, and __dirname/import.meta.url
path resolutions where tests read fixtures from disk.
- Drop tracked tsc-emit alongside engine .ts sources (auth-storage/logger/
skill-resolver/context-limit-detector/pi.{js,d.ts,*.map}). These were
accidentally committed in a merge and the stale pi.js was masking a real
test-mock vs source mismatch (tests imported "../pi.js" and vite preferred
the stale build over pi.ts).
- Add packages/engine/.gitignore to block future src/*.{js,d.ts,map}.
- Refactor plugin pi-module seams (openclaw/paperclip/hermes) to ESM-import
createFnAgent / promptWithFallback / describeModel from @fusion/engine
instead of require()-ing packages/engine/src/pi.js. Adds @fusion/engine to
the two plugin package.jsons that were missing it; exports describeModel
from the engine public API.
- Fix engine test mocks now that they run against current pi.ts: add
ModelRegistry.create static to mocks in pi.test.ts and pi-create-fn-agent
.test.ts; switch three boundary-result toEqual assertions to toMatchObject
so the new content/isError fields don't trip exact-match comparison.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The FN-2370 auto-resolved squash (7c1a1c36c) reverted three commits' worth
of refinements to the claude-cli provider because the branch contained
rebased duplicates of commits already on main. The auto-resolver picked
the older side line-by-line and dropped the newer.
Restored:
- /api/models filter logic (was inverted; emptied every model picker)
- Claude Opus 4.7 catalog entry in pi-claude-cli
- Provider card status text and toast messages (no longer claim a restart
is needed — the extension is always-loaded now)
- POST /api/auth/claude-cli returns restartRequired: false
Prevention:
- Regression tests on the /api/models useClaudeCli filter
- scripts/audit-squash-merge.mjs flags duplicate-cherry-pick risk and
touched-file overlap on any squash commit
- AGENTS.md documents the rebase-before-squash rule and requires the
merging agent to run the audit and triage every flagged item itself
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The pinned @mariozechner/pi-ai catalog tops out at Opus 4.6 / Sonnet 4.6
/ Haiku 4.5. Append a Claude Opus 4.7 entry (1M ctx, 128k max out,
$5/$25 per MTok) with id-dedupe so it becomes a no-op once the upstream
catalog catches up. Sonnet 4.6 and Haiku 4.5 remain current per
https://platform.claude.com/docs/en/about-claude/models/overview and
are already in the catalog, so no other additions are needed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- TaskCard: four catch((err: any) => err.message) promise handlers in
archive/unarchive/delete/move → catch((err) => getErrorMessage(err)).
- InlineCreateCard + QuickEntryBox: .catch((err: any)) model-load handlers
→ getErrorMessage(err) with existing @fusion/core import.
- TerminalModal: drop (navigator as any).maxTouchPoints — modern lib.dom
types already expose the property.
- serve.ts: remove unused any annotation on OpenRouter model mapper; the
array element type is already inferred from json.data.
- pi.js, runtime-resolution.ts, dashboard.ts, serve.ts, dev-server-port-
detect.ts, devserver-manager.ts: drop now-stale eslint-disable comments
that the cleanup made redundant.
Fix a prompt-builder regression surfaced by agent's `any` cleanup: toolCall
with a raw string `arguments` field must be preserved verbatim (JSON-quoted)
rather than coerced to `{}`; restores a previously-passing test.
Then promote @typescript-eslint/no-explicit-any from warn → error. Future
new anys must either come with a one-line disable + justification or use a
real type. Workspace is now lint-clean (0 problems).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Parallel subagent pass: four typescript-pro agents on non-overlapping scopes.
Patterns applied:
- catch (err: any) { ... err.message ... } → catch (err) { ... getErrorMessage(err) ... }
using the new @fusion/core helper. Bare catch {} where the error was unused.
- SQLite row types: defined typed XxxRow interfaces per table and cast
.all()/.get() results via `as unknown as XxxRow[]` (the double cast is
required because better-sqlite3 returns Record<string, SQLOutputValue>).
- rowToX(row: any) converters: typed argument with the matching row interface.
- Dynamic settings key writes: (settings as Record<string, unknown>)[key].
- React event handlers and setState callbacks: inferred types or concrete
React.{Mouse,Change,Form}Event<...> where needed.
- pi-claude-cli: local PiMessage / PiContext duck types to avoid re-typing
pi-ai concrete shapes; typed Claude stream event message fields.
72 files changed, ~400 anys eliminated. Typecheck passes across the workspace.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Widen config match from top-level *.cjs to **/*.cjs so nested CommonJS
scripts (e.g. mcp-schema-server.cjs) get Node globals and require() allowance.
- Replace lazy require() in claude-skills.ts with a normal top-level fs import.
- Tighten an any-typed tool map in pi-claude-cli to { name: string }.
Clears the remaining 12 lint errors; workspace now has 0 errors, 450 warnings.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the stray useClaudeCli settings checkbox + onboarding question
with a proper provider-card UX. The card lives next to OAuth + API-key
cards in onboarding and settings, with Enable/Disable + Test actions.
Backend:
- Vendors rchern/pi-claude-cli@0.3.1 as packages/pi-claude-cli
(MIT, attribution in UPSTREAM.md). Lets us bump peer-dep on
pi-coding-agent in lockstep with Fusion (upstream pinned ^0.52.0
vs ours ^0.62.0) and fix bugs without waiting on upstream.
- Adds @fusion/pi-claude-cli as a workspace dep of @runfusion/fusion
so users don't have to `npm install -g pi-claude-cli` manually.
- serve/daemon/dashboard conditionally load the extension via
discoverAndLoadExtensions() when GlobalSettings.useClaudeCli is on;
no side-effects on user ~/.fusion/agent/settings.json.
- New GET /api/providers/claude-cli/status: claude --version probe
+ toggle state + cached extension resolution.
- New POST /api/auth/claude-cli: flips useClaudeCli, refuses if the
claude binary is missing, fires the existing skill-backfill hook.
- /api/auth/status now injects a synthetic {id:"claude-cli", type:"cli"}
provider entry so onboarding + settings see a consistent list.
Frontend:
- New ClaudeCliProviderCard component shared between ModelOnboardingModal
and SettingsModal's Authentication section.
- New AuthProvider.type = "cli" variant.
- Removed the old "Route AI calls through the Claude CLI" checkbox from
Global Models settings and the opt-in step from the onboarding wizard.
- ProviderIcon gets a composite Anthropic-mark-plus-terminal glyph for
the claude-cli provider id.
Tests:
- 8 unit tests for extension resolution (@fusion/pi-claude-cli is
workspace-linked so these run in-tree).
- 2 unit tests for the binary probe.
- Existing /auth/status tests filter out the new synthetic entry so
they keep asserting structural OAuth/API-key behavior in isolation.
- The vendored package's own 296 tests still pass unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
