import type { AgencyConfig } from "../../config.js";
import { type EvalRecordExtractor, type EvalInputRunner } from "../../eval/runEvalInput.js";
import type { EvalRunResult, Input } from "../../eval/runTypes.js";
export type EvalRunCliOptions = {
    agent: string;
    inputs?: string;
    goal?: string;
    runId?: string;
    runsDir?: string;
    continueOnError?: boolean;
    verbose?: boolean;
    config?: AgencyConfig;
};
export type EvalRunLoadedInputsOptions = {
    agent: string;
    inputs: Input[];
    inputsSource: string;
    runId?: string;
    runsDir?: string;
    continueOnError?: boolean;
    verbose?: boolean;
    config?: AgencyConfig;
    /** Suppress compile progress lines for the agent compile.
     *  (Currently unused — kept for API compatibility.) */
    quietCompile?: boolean;
    /** Pipe agent subprocess stdout/stderr through to the console. Defaults to true. */
    pipeAgentOutput?: boolean;
    /**
     * Workdir seed override. When set, used verbatim for every input — no
     * closure-base derivation, no relpath computation. The optimizer always
     * sets this from `source.baseDir` + `source.entryFile`, so the seed
     * cannot silently diverge from the closure walk.
     *
     * Mutually exclusive with per-input `working_dir`: setting both is an
     * error.
     */
    seed?: {
        dir: string;
        agentRelPath: string;
    };
    /** Candidate-file overlay applied to every input in this call (over the
     *  seeded copy, before compile). Plain eval never sets this. */
    overlayFiles?: Record<string, string>;
};
export declare function resolveEvalRunTarget(target: string): {
    agentFile: string;
    node: string;
    label: string;
};
export declare function validateInputSelection(opts: {
    inputs?: string;
    goal?: string;
}): "inputs" | "goal";
/**
 * Run a compiled agent against an input suite and write per-input artifacts.
 *
 * The orchestration here is intentionally tiny: build the run state, loop
 * through inputs calling `runEvalInput` (the shared boundary), write a
 * summary. All per-input error handling lives inside `runEvalInput`.
 */
export declare function evalRun(opts: EvalRunCliOptions, overrides?: {
    runner?: EvalInputRunner;
    extractor?: EvalRecordExtractor;
}): Promise<EvalRunResult>;
export declare function evalRunLoadedInputs(opts: EvalRunLoadedInputsOptions, overrides?: {
    runner?: EvalInputRunner;
    extractor?: EvalRecordExtractor;
}): Promise<EvalRunResult>;
/**
 * Extractor for the optimizer: grades the entry node's return value (not the
 * last LLM completion) when `evalOutput()` wasn't called, and omits the
 * evalValue/evalOutput "did you forget to call…" warnings — inputs come from
 * the input spec and the graded output is the return value, so neither applies.
 */
export declare const optimizeEvalRecordExtractor: EvalRecordExtractor;