export interface CodeMetadata {
    id: string;
    filePath: string;
    functionName?: string;
    className?: string;
    language: string;
    complexity: number;
    lineCount: number;
    lastModified: Date;
}
export interface SemanticSearchResult {
    id: string;
    code: string;
    metadata: CodeMetadata;
    similarity: number;
}
export declare class SemanticVectorDB {
    private db;
    private initialized;
    private localEmbeddingPipeline;
    private embeddingCache;
    private readonly EMBEDDING_CACHE_SIZE;
    private readonly LOCAL_EMBEDDING_DIMENSION;
    private hasLoggedEmbeddingStart;
    constructor(_apiKey?: string);
    /**
     * Initialize local embedding pipeline using transformers.js
     */
    private initializeLocalEmbeddings;
    initialize(collectionName?: string): Promise<void>;
    storeCodeEmbedding(code: string, metadata: CodeMetadata): Promise<void>;
    storeMultipleEmbeddings(codeChunks: string[], metadataList: CodeMetadata[]): Promise<void>;
    findSimilarCode(query: string, limit?: number, filters?: Record<string, any>): Promise<SemanticSearchResult[]>;
    findSimilarCodeByFile(filePath: string, limit?: number): Promise<SemanticSearchResult[]>;
    findSimilarCodeByLanguage(query: string, language: string, limit?: number): Promise<SemanticSearchResult[]>;
    updateCodeEmbedding(id: string, code: string, metadata: CodeMetadata): Promise<void>;
    deleteCodeEmbedding(id: string): Promise<void>;
    deleteCodeEmbeddingsByFile(filePath: string): Promise<void>;
    getCollectionStats(): Promise<{
        count: number;
        metadata: any;
    }>;
    private generateEmbedding;
    /**
     * Generate real semantic embeddings using local method
     */
    private generateRealSemanticEmbedding;
    /**
     * Get local embeddings using transformers.js or fallback method
     */
    private getLocalEmbedding;
    /**
     * Generate advanced local semantic embeddings using multiple techniques
     */
    private generateAdvancedLocalEmbedding;
    /**
     * For backward compatibility - use the proper local embedding method
     */
    private generateLocalEmbedding; /**
     * Extract structural code features
     */
    private extractStructuralFeatures;
    /**
     * Extract semantic token features
     */
    private extractSemanticFeatures;
    /**
     * Extract AST-based features
     */
    private extractASTFeatures;
    /**
     * Extract contextual features
     */
    private extractContextFeatures;
    /**
     * Extract meaningful programming tokens
     */
    private extractMeaningfulTokens;
    /**
     * Get programming-specific vocabulary
     */
    private getProgrammingVocabulary;
    /**
     * Calculate token frequency
     */
    private calculateTokenFrequency;
    /**
     * Preprocess code for embedding
     */
    private preprocessCodeForEmbedding;
    /**
     * Create cache key from code
     */
    private createCacheKey;
    /**
     * Cache embedding with LRU eviction
     */
    private cacheEmbedding;
    /**
     * Normalize vector for cosine similarity
     */
    private normalizeVector;
    private getVocabulary;
    close(): Promise<void>;
}
//# sourceMappingURL=vector-db.d.ts.map