Skip to content

EvalLlmJudgeAssertion

PromptScript API


Defined in: core/src/types/eval.ts:171

Passes when the majority of judge runs say pass. The rubric receives the inputs response, case_inputs, expected, criteria and the reserved boundary.

Extends

Properties

criteria?

optional criteria?: string

Defined in: core/src/types/eval.ts:179

Extra criteria passed to the rubric as its criteria input


field?

optional field?: string

Defined in: core/src/types/eval.ts:109

Output property the assertion reads; absent means the whole response text

Inherited from

EvalAssertionBase.field


id

id: string

Defined in: core/src/types/eval.ts:107

Assertion id, unique in the suite.

Pattern

^[A-Za-z0-9][A-Za-z0-9_.-]{0,63}$

Example

"decision-matches"

Inherited from

EvalAssertionBase.id


kind

kind: "llm-judge"

Defined in: core/src/types/eval.ts:172


model?

optional model?: string

Defined in: core/src/types/eval.ts:181

Model reference the judge runs on; absent means the suite judge, then the rubric’s own model


rubric

rubric: string

Defined in: core/src/types/eval.ts:177

Prompt id of the rubric.

Example

"acme-support/judge-polite-reply"

runs?

optional runs?: EvalJudgeRuns

Defined in: core/src/types/eval.ts:183

Judge runs per case; absent means the suite judge setting