RatingConfig

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:644

Configuration for review rating feature. Allows configuring automated review scoring with pass/fail thresholds.

Properties

enabled?

optional enabled?: boolean

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:649

Enable or disable review rating.

Default

true

---

errorOnReviewFail?

optional errorOnReviewFail?: boolean

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:670

Exit with error code 1 when review fails (below threshold). When false, exits normally (code 0) regardless of rating.

Default

true

---

maxRating?

optional maxRating?: number

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:659

Highest allowed value on the rating scale.

Default

10

---

minRating?

optional minRating?: number

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:664

Lowest allowed value on the rating scale.

Default

0

---

passThreshold?

optional passThreshold?: number

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:654

Minimum score (0-10) required to pass the review.

Default

6

---

timeoutMs?

optional timeoutMs?: number

Defined in: _worktrees/docs-release/gaunt-sloth/packages/core/src/config/types.ts:689

Wall-clock budget (ms) for the single rating model call.

GS2-105 — rating is a SECOND model call fired from afterAgent once the review prose is already written, and it renders nothing while it runs. Without a bound, a model that fails to emit the rating tool call generates prose until its context window stops it: measured at 15,385 tokens and 485 seconds on gemma4:12b, against a healthy rating call of ~50 seconds. The budget turns that silent hang into a bounded, legible failure, after which the existing missing-artifact path reports it and exits non-zero.

It is a WALL-CLOCK budget for one call, not a per-token cap, because the per-token knob is not portable: numPredict is a constructor field on ChatOllama rather than a call option, and that provider declares tool_choice?: never, so neither of the obvious source-level remedies is available where this actually bites. timeout is a standard RunnableConfig field, so it works on every provider and genuinely aborts the request instead of merely abandoning it.

Default

120000