/** * AI quality evaluator for implementation comparisons. * * Runs a single-turn, non-agentic AI invocation via {@link AIService} * to blindly evaluate two implementations. Randomizes presentation * order to prevent position bias. * * @module */ import type { AIService } from '../ai/service.js'; import type { ImplementationEvaluation } from './types.js'; /** * Evaluate two implementations blindly using an AI model. * * Randomizes which implementation is "Plan A" vs "Plan B" to prevent * position bias, then maps scores back to the original labels. * * @param task - The original task description * @param withDocsImpl - Implementation log from the "with docs" run * @param withoutDocsImpl - Implementation log from the "without docs" run * @param aiService - AI service instance for subprocess management * @param model - Model to use for evaluation * @param debug - Enable debug logging * @returns Evaluation result, or null if evaluation fails */ export declare function evaluateImplementations(task: string, withDocsImpl: string, withoutDocsImpl: string, aiService: AIService, model?: string, debug?: boolean): Promise; //# sourceMappingURL=evaluator.d.ts.map