EvaluatorRatingScale Class
Represents a rating scale for custom LLM-as-a-Judge evaluators.
Rating scales define how the evaluator scores agent performance. Use either categorical (discrete labels) or numerical (labeled numeric values) scales.
Example
ts
// Categorical rating scale
const categorical = agentcore.EvaluatorRatingScale.categorical([
{ label: 'Good', definition: 'The response fully addresses the query.' },
{ label: 'Bad', definition: 'The response fails to address the query.' },
]);
// Numerical rating scale
const numerical = agentcore.EvaluatorRatingScale.numerical([
{ label: 'Poor', definition: 'Inadequate response.', value: 1 },
{ label: 'Good', definition: 'Adequate response.', value: 3 },
{ label: 'Excellent', definition: 'Outstanding response.', value: 5 },
]);
Methods
categorical(options: CategoricalRatingOption[]): EvaluatorRatingScale // chainable@staticmethod def categorical(options: typing.List[CategoricalRatingOption]) -> EvaluatorRatingScalepublic static EvaluatorRatingScale categorical(List<CategoricalRatingOption> options)public static EvaluatorRatingScale Categorical(CategoricalRatingOption[] options)func Categorical(Options *[]CategoricalRatingOption) EvaluatorRatingScalestatic#Creates a categorical rating scale.
Categorical scales define discrete labels for scoring, such as "Good" / "Bad" or "Pass" / "Fail".
1 parameter
Parameters
optionsoptionsoptionsoptionsOptions— - The categorical rating options (at least 1 required).
numerical(options: NumericalRatingOption[]): EvaluatorRatingScale // chainable@staticmethod def numerical(options: typing.List[NumericalRatingOption]) -> EvaluatorRatingScalepublic static EvaluatorRatingScale numerical(List<NumericalRatingOption> options)public static EvaluatorRatingScale Numerical(NumericalRatingOption[] options)func Numerical(Options *[]NumericalRatingOption) EvaluatorRatingScalestatic#Creates a numerical rating scale.
Numerical scales define labeled numeric values for scoring, such as 1 (Poor) through 5 (Excellent).
1 parameter
Parameters
optionsoptionsoptionsoptionsOptions— - The numerical rating options (at least 1 required).