agentfootprint/dist/esm/classify/mock.d.ts

The explainable agent framework — backtrack a wrong answer to the exact context that caused it (evidence, not guesses). Built on footprintjs.

/**
 * classify/mock — a scripted `Classifier` for tests and the bench.
 *
 * Pattern: the `mock()` LLM provider's shape — a script, replayed in order,
 *          or a function of the request; every request recorded on `calls`.
 * Role:    the one classifier the suite and `bench/findings-shuffle.mjs`
 *          may run without a key. It produces exactly what its script says
 *          (a scripted distribution is still the SCRIPT's data, never the
 *          library's guess) and throws exactly what the script throws, so a
 *          failure path is testable to the byte.
 */
import { type Classifier, type ClassifyRequest, type ClassifyResult } from './types.js';
/** A classifier that answers from a script and records what it was asked. */
export interface MockClassifier extends Classifier {
    /** Every request, in call order — plain references to what the caller passed. */
    readonly calls: readonly ClassifyRequest[];
}
export type MockClassifierScript = readonly ClassifyResult[] | ((request: ClassifyRequest, index: number) => ClassifyResult | Promise<ClassifyResult>);
/**
 * `mockClassifier([r1, r2])` answers r1 then r2 and refuses a third call
 * (`ClassifierError`, so the exhaustion is a failure row, never a repeat);
 * `mockClassifier((req, i) => …)` computes each answer, and may throw to
 * script a failure. `name` is `'mock'`; the result's `model` is the
 * script's own.
 */
export declare function mockClassifier(script: MockClassifierScript): MockClassifier;