/**
 * SpecVectorIndex
 *
 * Builds and queries a LanceDB vector index over OpenSpec spec files.
 * Each record represents a logical section (requirement, purpose, design note...)
 * parsed from the Markdown spec files.
 *
 * Storage: <outputDir>/vector-index/  (same LanceDB folder as VectorIndex, different table)
 * Table name: "specs"
 *
 * Usage:
 *   // Build (or rebuild)
 *   await SpecVectorIndex.build(outputDir, specsDir, mappingJson, embedSvc);
 *
 *   // Search
 *   const results = await SpecVectorIndex.search(outputDir, "email validation", embedSvc);
 */
import type { EmbeddingService } from './embedding-service.js';
export interface SpecRecord {
    id: string;
    domain: string;
    section: string;
    title: string;
    /** Concatenated text used for embedding */
    text: string;
    /** Source files linked to this requirement (from mapping.json) */
    linkedFiles: string;
    /** Embedding vector */
    vector: number[];
}
export interface SpecSearchResult {
    record: {
        id: string;
        domain: string;
        section: string;
        title: string;
        text: string;
        /** Parsed from JSON -- list of source file paths linked to this spec section */
        linkedFiles: string[];
    };
    score: number;
}
export declare class SpecVectorIndex {
    /**
     * Discover all spec.md files under specsDir, parse them, enrich with mapping,
     * embed, and write to LanceDB table "specs".
     *
     * @param outputDir  Path to .openlore/analysis/
     * @param specsDir   Path to openspec/specs/ (or any directory containing domain/spec.md files)
     * @param mappingJsonPath  Path to mapping.json (optional -- no enrichment if absent)
     */
    static build(outputDir: string, specsDir: string, embedSvc: EmbeddingService | null, mappingJsonPath?: string, decisionsDir?: string): Promise<{
        recordCount: number;
        hasEmbeddings: boolean;
    }>;
    /**
     * Semantic search over the spec index.
     */
    static search(outputDir: string, query: string, embedSvc: EmbeddingService | null | undefined, opts?: {
        limit?: number;
        domain?: string;
        section?: string;
    }): Promise<SpecSearchResult[]>;
    /**
     * BM25-only search over the spec corpus: used when no embedding service is
     * available or the index was built without embeddings. Scores the full
     * corpus with BM25 and returns the top `limit` matching sections.
     */
    private static _bm25Only;
    /**
     * Returns true if the spec index table exists.
     */
    static exists(outputDir: string): boolean;
}
//# sourceMappingURL=spec-vector-index.d.ts.map