Home > intention-kernel/testing > runEvaluation

runEvaluation() function

Execute isolated multi-turn cases using an application-owned adapter.

Signature:

export declare function runEvaluation(options: EvaluationOptions): Promise<EvaluationReport>;

Parameters

Parameter	Type	Description
options	EvaluationOptions

Returns:

Promise<EvaluationReport>

Complete case inventory with separate passed, failed, error, blocked and skipped counts.

Exceptions

EvaluationError for invalid inputs, incompatible resume or a failed observer.

Remarks

Every transition emits a deeply immutable checkpoint before its event. Unchanged case evidence retains object identity across checkpoints. Cases run concurrently; exchanges within a case remain sequential. A failed expectation does not stop other cases. No request is automatically retried. Cancellation and deadlines cannot undo an external effect; inspect receipts before retrying it. Resume requires an identical suite/selection/target/adapter fingerprint. Only adapters declaring supportsResume may resume an incomplete conversation prefix.