/**
 * Project Mind MCP - Embeddings Engine
 *
 * Local embeddings using ONNX for semantic search.
 * Uses all-MiniLM-L6-v2 model for high-quality, fast embeddings.
 *
 * Key features:
 * - Local inference (no API calls)
 * - ~50ms per query
 * - 384-dimensional embeddings
 * - Automatic caching of model
 */
/**
 * Generate embedding for a single text.
 *
 * @param text - Text to embed
 * @returns 384-dimensional embedding as Float32Array
 */
export declare function embed(text: string): Promise<Float32Array>;
/**
 * Generate embeddings for multiple texts.
 *
 * @param texts - Array of texts to embed
 * @returns Array of 384-dimensional embeddings
 */
export declare function embedBatch(texts: string[]): Promise<Float32Array[]>;
/**
 * Calculate cosine similarity between two embeddings.
 * Both vectors should already be normalized.
 *
 * @param a - First embedding
 * @param b - Second embedding
 * @returns Similarity score between -1 and 1
 */
export declare function cosineSimilarity(a: Float32Array, b: Float32Array): number;
/**
 * Find most similar embeddings from a collection.
 *
 * @param queryEmbedding - The query embedding
 * @param candidates - Array of candidates with id and embedding
 * @param limit - Maximum results to return
 * @param minSimilarity - Minimum similarity threshold
 * @returns Sorted array of matches with similarity scores
 */
export declare function findSimilar(queryEmbedding: Float32Array, candidates: Array<{
    id: string | number;
    embedding: Float32Array;
}>, limit?: number, minSimilarity?: number): Array<{
    id: string | number;
    similarity: number;
}>;
/**
 * Serialize embedding to Buffer for database storage.
 */
export declare function serializeEmbedding(embedding: Float32Array): Buffer;
/**
 * Deserialize embedding from database storage.
 */
export declare function deserializeEmbedding(buffer: Buffer): Float32Array;
/**
 * Get embedding dimension (for schema validation).
 */
export declare function getEmbeddingDimension(): number;
/**
 * Check if embeddings engine is ready.
 */
export declare function isReady(): boolean;
/**
 * Preload the model (call at startup for faster first query).
 */
export declare function preload(): Promise<void>;
//# sourceMappingURL=embeddings.d.ts.map