memory-engineering-mcp/dist/embeddings/codeEmbeddings.d.ts

🧠 AI Memory System powered by MongoDB Atlas & Voyage AI - Autonomous memory management with zero manual work

import type { CodeChunk } from '../types/memory-v5.js';
/**
 * Generate embeddings for code chunks using voyage-code-3 model.
 * Each chunk is embedded with its signature, content, and metadata.
 */
export declare function generateCodeEmbeddings(chunks: Omit<CodeChunk, '_id' | 'contentVector' | 'createdAt' | 'updatedAt'>[]): Promise<number[][]>;
/**
 * Generate embedding for a code search query using voyage-3 model
 */
export declare function generateCodeQueryEmbedding(query: string): Promise<number[]>;
/**
 * Example of how contextualized embeddings solve the "golden chunk" problem:
 *
 * File: userService.ts
 * Chunks:
 * 1. "import { User } from './models/User';"
 * 2. "export class UserService { ... }"
 * 3. "async authenticate(email: string, password: string) { ... }"
 *
 * With standard embeddings:
 * - Chunk 3 loses context about UserService class
 * - Search for "UserService authenticate" might not find chunk 3
 *
 * With contextualized embeddings:
 * - Chunk 3 is embedded knowing it's part of UserService
 * - Search for "UserService authenticate" correctly finds chunk 3
 * - Each chunk "knows" about the imports and class structure
 */ 
//# sourceMappingURL=codeEmbeddings.d.ts.map