Skip to content

docs(config): document the reviewer / debate / tournament knobs - #317

Open
Jacky628 wants to merge 1 commit into
aiming-lab:mainfrom
Jacky628:docs/document-review-decoupling-knobs
Open

docs(config): document the reviewer / debate / tournament knobs#317
Jacky628 wants to merge 1 commit into
aiming-lab:mainfrom
Jacky628:docs/document-review-decoupling-knobs

Conversation

@Jacky628

Copy link
Copy Markdown
Contributor

Follow-up to #316, as requested: "Opt-in features that aren't advertised tend to stay unused."

Documents the nine settings in config.researchclaw.example.yaml. The reviewer_* fields are written out explicitly with empty values (empty = today's behaviour, so the file stays a working config); debate_* and tournament_* are commented out.

Scope is what's actually wired up on main, which is narrower than the engines are capable of — debate and the tournament are documented for Stage 8 only, and there is no reasoning_effort setting. I checked the wiring rather than describing the engines' full capability, so the file shouldn't promise anything that isn't there.

Beyond listing the fields, it records the parts a user cannot infer from the names:

  • the tournament takes priority over debate on Stage 8, so debate_enabled: true alone does nothing while the tournament is also on;
  • a debate panel needs >= 2 distinct models after deduplication, otherwise Stage 8 quietly falls back to the single-model path;
  • the four ARC_* env overrides, including the two ablation switches.

Verified by parsing the example config through RCConfig:

reviewer_model        = ''
debate_enabled        = False
tournament_enabled    = False
debate_rounds         = 1
tournament_candidates = 3

The documented defaults match what the loader actually produces.

No code changes. Full suite: 2951 passed, 0 failed, 56 skipped.

🤖 Generated with Claude Code

Follow-up to aiming-lab#316: opt-in features that aren't advertised tend to stay
unused. Documents the nine settings in the example config, with the
reviewer fields written out explicitly (empty = current behaviour) and
the debate/tournament ones commented out.

Scope is limited to what is actually wired up on main, which is narrower
than the engines can do: debate and the tournament are documented for
Stage 8 only, and there is no reasoning_effort setting. Also records the
non-obvious parts a user cannot infer from field names — that the
tournament takes priority over debate on Stage 8, that a debate panel
needs >= 2 distinct models after deduplication or it silently falls back
to the single-model path, and the four ARC_* env overrides.

Verified by parsing the example config through RCConfig: reviewer_model
empty, debate_enabled and tournament_enabled False — the documented
defaults match what the loader produces.

No code changes. Full suite: 2951 passed, 0 failed, 56 skipped.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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.

1 participant