import { type GraphStructureMetrics } from '../pipeline/analyze.js';
import { type CorpusBaselineSource } from './benchmark/corpus.js';
import { type BenchmarkPromptExecution, type BenchmarkPromptRunnerResult } from './benchmark/runner.js';
import { type BenchmarkMissingExpectedLabels, type BenchmarkQuestionInput, type BenchmarkQuestionResult } from './benchmark/questions.js';
export { loadBenchmarkQuestions, querySubgraphTokens, type BenchmarkQuestionInput } from './benchmark/questions.js';
export declare const SAMPLE_QUESTIONS: string[];
export interface BenchmarkSuccessResult {
    corpus_tokens: number;
    corpus_words: number;
    corpus_source: CorpusBaselineSource;
    nodes: number;
    edges: number;
    structure_signals: GraphStructureMetrics | null;
    question_count: number;
    matched_question_count: number;
    unmatched_questions: string[];
    expected_label_count: number;
    matched_expected_label_count: number;
    missing_expected_labels: BenchmarkMissingExpectedLabels[];
    avg_query_tokens: number;
    avg_effective_query_tokens?: number;
    avg_reused_context_tokens?: number;
    avg_total_tokens?: number | null;
    reduction_ratio: number;
    effective_reduction_ratio?: number;
    provider_proof?: {
        input_tokens_basis: 'provider_reported' | 'mixed' | 'estimated';
        effective_tokens_basis: 'provider_cache_read_tokens' | 'provider_input_minus_zero_cache' | 'mixed' | 'session_reuse_estimate';
        total_tokens_basis: 'provider_reported' | 'mixed' | 'not_available';
        usage_runs: number;
        total_runs: number;
        providers: string[];
    };
    per_question: BenchmarkQuestionResult[];
}
export interface BenchmarkErrorResult {
    error: string;
}
export type BenchmarkResult = BenchmarkSuccessResult | BenchmarkErrorResult;
export interface BenchmarkRunOptions {
    execTemplate?: string;
    outputDir?: string;
    now?: Date;
    retrievalBudget?: number;
    runner?: (execution: BenchmarkPromptExecution) => Promise<BenchmarkPromptRunnerResult>;
}
export declare function runBenchmark(graphPath?: string, corpusWords?: number | null, questions?: BenchmarkQuestionInput[], options?: BenchmarkRunOptions): BenchmarkResult | Promise<BenchmarkResult>;
export declare function printBenchmark(result: BenchmarkResult): void;