@mastra/core/dist/datasets/experiment/executor.d.ts

import type { Agent } from '../../agent/index.js';
import type { MastraScorer } from '../../evals/base.js';
import type { ScorerRunInputForAgent, ScorerRunOutputForAgent } from '../../evals/types.js';
import type { VersionOverrides } from '../../mastra/types.js';
import type { TargetType } from '../../storage/types.js';
import type { StepResult, Workflow } from '../../workflows/index.js';
import type { ItemToolMock, ToolMockReport } from './tool-mocks.js';
/**
 * Target types supported for dataset execution.
 * Agent and Workflow are Phase 2; scorer and processor are Phase 4.
 */
export type Target = Agent | Workflow | MastraScorer<any, any, any, any>;
/**
 * Result from executing a target against a dataset item.
 */
export interface ExecutionResult {
    /** Output from the target (null if failed) */
    output: unknown;
    /** Structured error if execution failed */
    error: {
        message: string;
        stack?: string;
        code?: string;
    } | null;
    /** Trace ID from agent/workflow execution (null for scorers or errors) */
    traceId: string | null;
    /** Root span ID from agent/workflow execution (null when not traced) */
    spanId?: string | null;
    /** Structured input for scorers (extracted from agent scoring data) */
    scorerInput?: ScorerRunInputForAgent;
    /** Structured output for scorers (extracted from agent scoring data) */
    scorerOutput?: ScorerRunOutputForAgent;
    /** Per-step results from a workflow run, keyed by step ID */
    stepResults?: Record<string, StepResult<any, any, any, any>>;
    /** Order in which workflow steps actually executed */
    stepExecutionPath?: string[];
    /** Diagnostic receipt for item-level tool mocks (agent targets only) */
    toolMockReport?: ToolMockReport;
}
/**
 * Execute a dataset item against a target (agent, workflow, scorer, processor).
 * Phase 2: agent/workflow. Phase 4: scorer. Processor deferred.
 */
export declare function executeTarget(target: Target, targetType: TargetType, item: {
    input: unknown;
    groundTruth?: unknown;
    metadata?: Record<string, unknown>;
    resumeSteps?: Record<string, unknown>;
    resumeData?: unknown;
}, options?: {
    signal?: AbortSignal;
    requestContext?: Record<string, unknown>;
    experimentId?: string;
    versions?: VersionOverrides;
    /** Item-level static tool mocks (agent targets only). */
    toolMocks?: ItemToolMock[];
}): Promise<ExecutionResult>;
//# sourceMappingURL=executor.d.ts.map