import { KnowledgeGraph } from '../../contracts/graph.js';
import type { ContextSessionDiagnostics } from '../../contracts/context-session.js';
import { type PromptRunnerUsage } from '../prompt-runner.js';
import { type BenchmarkPromptArtifacts, type BenchmarkPromptTokenSource } from './runner.js';
export interface BenchmarkQuestionResult {
    id?: string;
    description?: string;
    question: string;
    query_tokens: number;
    effective_query_tokens?: number;
    reused_context_tokens?: number;
    session_diagnostics?: ContextSessionDiagnostics;
    reduction: number;
    expected_labels: string[];
    matched_expected_labels: string[];
    missing_expected_labels: string[];
    total_tokens?: number | null;
    prompt_tokens_estimated?: number;
    prompt_token_source?: BenchmarkPromptTokenSource;
    usage?: PromptRunnerUsage | null;
    answer_text?: string | null;
    elapsed_ms?: number;
    artifacts?: BenchmarkPromptArtifacts;
}
export interface BenchmarkQuestionSpec {
    id?: string;
    description?: string;
    question: string;
    expected_labels?: string[];
}
export type BenchmarkQuestionInput = string | BenchmarkQuestionSpec;
export interface BenchmarkMissingExpectedLabels {
    question: string;
    labels: string[];
}
export interface BenchmarkQuestionEvaluation {
    question: string;
    result: BenchmarkQuestionResult | null;
    expected_label_count: number;
    matched_expected_label_count: number;
    missing_expected_labels: BenchmarkMissingExpectedLabels | null;
}
export declare function normalizeExpectedLabel(label: string): string;
export declare function normalizeBenchmarkQuestion(question: BenchmarkQuestionInput): BenchmarkQuestionSpec;
export declare function loadBenchmarkQuestions(questionsPath: string): BenchmarkQuestionSpec[];
export declare function querySubgraphTokens(graph: KnowledgeGraph, question: string, depth?: number): number;
export declare function evaluateBenchmarkQuestion(graph: KnowledgeGraph, question: BenchmarkQuestionInput, corpusTokens: number, depth?: number): BenchmarkQuestionEvaluation;