Skip to content

feat(personify): connect the dormant analyst→steer wire + registryScopeAnalyst#284

Merged
drewstone merged 1 commit into
mainfrom
feat/analyst-steer-wire
Jun 14, 2026
Merged

feat(personify): connect the dormant analyst→steer wire + registryScopeAnalyst#284
drewstone merged 1 commit into
mainfrom
feat/analyst-steer-wire

Conversation

@drewstone

Copy link
Copy Markdown
Contributor

Connects the analyst→steer seam that was wired to a hardcoded empty array (loopUntil called spec.until(state, []), widen called spec.gate.decide(s, [], …); createScopeAnalyst had zero live callers).

  • ShapeContext.analyst? + RunPersonifiedOptions.analyst? (threaded via createShapeContext/runPersonified).
  • loopUntil/widen: const findings = ctx.analyst ? await ctx.analyst.analyze({task, settledSoFar, nodeId: scope.view.root}) : [] → fed to the gate (was []). Dormant [] default preserved (no behavior change without an analyst).
  • registryScopeAnalyst(registry, buildInputs) — the panel-of-analysts adapter: merges N analyst KINDS via the existing AnalystRegistry.run, routed through the existing assertTraceDerivedFindings firewall (single-sourced). NOT panel() (that's judges-vs-artifact).
  • Equal-k preserved: the analyst spawns into the same scope (unchanged createScopeAnalyst), conserved-pool-charged.
  • Hazard documented: widen's streaming cursor + shared-cursor analyst drain would swallow a sibling settlement on a non-flat gate; the shipped flatWidenGate default never widens so the wire is exact, and the hazard is in the docstring (not papered over).

+7 tests (876 pass total), tsc + Biome clean. The shared prerequisite for the remaining analyst features (online intra-turn, long-range) + the intelligence-SDK hookup.

…peAnalyst

Thread an optional ScopeAnalyst<D> through ShapeContext / RunPersonifiedOptions
/ createShapeContext, and consult it at the two combinator steer gates
(loopUntil.until, widen.gate.decide) instead of the hardcoded empty findings
array. The analyst spawns into the SAME scope (createScopeAnalyst), so its
compute stays on the conserved budget pool — equal-k holds by construction.
Absent an analyst the gates keep the dormant [] default (no behavior change).

Add registryScopeAnalyst: a ScopeAnalyst backed by an AnalystRegistryLike that
merges N analyst kinds into one findings list and pipes them through the same
assertTraceDerivedFindings firewall createScopeAnalyst uses (single-sourced
selector≠judge). The ScopeAnalyzeInput→registry-inputs projection is
caller-supplied (buildInputs), not an invented bridge.

@tangletools tangletools left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

✅ Auto-approved PR — 5b6321ab

Blanket team auto-approval is enabled for this reviewer service.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

tangletools · auto-approval · reason: blanket_auto_approve · 2026-06-14T00:49:45Z

@drewstone drewstone merged commit 1ca53f8 into main Jun 14, 2026
1 check passed
drewstone added a commit that referenced this pull request Jun 14, 2026
- feat(conversation): runPersonaConversation + runPersonaDispatch — the persona
  loop runner; any AgentProfile evaluated as a multi-round conversation, drops
  into runProfileMatrix as dispatch (#282)
- feat(personify): connect the dormant analyst→steer wire + registryScopeAnalyst (#284)
- feat(skills): build-with-agent-runtime canonical spine (#285)
- fix(tool-loop): strict-model tool-call history (#286)
- docs: canonical-api.md API reference (#283)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants