hikma-engine/dist/modules/embedding-service.d.ts

Code Knowledge Graph Indexer - A sophisticated TypeScript-based indexer that transforms Git repositories into multi-dimensional knowledge stores for AI agents

/**
 * @file Responsible for generating vector embeddings for various node types.
 */
import { BaseNode, NodeWithEmbedding } from '../types';
import { ConfigManager } from '../config';
export declare class EmbeddingService {
    private config;
    private logger;
    private model;
    /**
     * Initializes the Embedding Service.
     * @param {ConfigManager} config - Configuration manager instance.
     */
    constructor(config: ConfigManager);
    /**
     * Loads the pre-trained embedding model.
     */
    loadModel(): Promise<void>;
    /**
     * Extracts meaningful text content from a node for embedding generation.
     * @param {BaseNode} node - The node to extract text from.
     * @returns {string} The extracted text content.
     */
    private getTextForNode;
    /**
     * Generates a vector embedding for a given text.
     * @param {string} text - The text to embed.
     * @param {boolean} isQuery - Whether this text is a search query (requires special prompt for some models).
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    private generateEmbedding;
    private simpleHash;
    /**
     * Normalizes user-provided OpenAI-compatible API URL to a base URL.
     * Accepts inputs like base or full paths and returns a clean base such as http://localhost:11434
     */
    private normalizeOpenAIBaseUrl;
    /**
     * Tests connection to LM Studio server.
     */
    private testLMStudioConnection;
    /**
     * Generates embedding using transformers.js pipeline.
     * @param {string} text - The text to generate an embedding for.
     * @param {boolean} isQuery - Whether this text is a search query.
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    private generateTransformersEmbedding;
    /**
     * Generates embedding using Python script.
     * @param {string} text - The text to generate an embedding for.
     * @param {boolean} isQuery - Whether this text is a search query.
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    private generatePythonEmbedding;
    /**
     * Generates embedding using LM Studio server.
     * @param {string} text - The text to generate an embedding for.
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    private generateLMStudioEmbedding;
    /**
     * Generates embedding using OpenAI-compatible API (like Ollama).
     * @param {string} text - The text to generate an embedding for.
     * @param {boolean} isQuery - Whether this is a query embedding.
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    private generateOpenAIEmbedding;
    /**
     * Processes nodes in batches to generate embeddings efficiently.
     * @param {BaseNode[]} nodes - Array of nodes to embed.
     * @returns {Promise<NodeWithEmbedding[]>} Array of nodes with embeddings.
     */
    private processBatch;
    /**
     * Generates embeddings for all provided nodes.
     * @param {BaseNode[]} nodes - Array of nodes to generate embeddings for.
     * @returns {Promise<NodeWithEmbedding[]>} Array of nodes with embeddings attached.
     */
    embedNodes(nodes: BaseNode[]): Promise<NodeWithEmbedding[]>;
    /**
     * Generates an embedding for a single text query (useful for search).
     * @param {string} query - The query text to embed.
     * @returns {Promise<number[]>} The embedding vector for the query.
     */
    embedQuery(query: string): Promise<number[]>;
    /**
     * Calculates cosine similarity between two embedding vectors.
     * @param {number[]} embedding1 - First embedding vector.
     * @param {number[]} embedding2 - Second embedding vector.
     * @returns {number} Cosine similarity score between -1 and 1.
     */
    calculateSimilarity(embedding1: number[], embedding2: number[]): number;
    /**
     * Finds the most similar nodes to a query embedding.
     * @param {number[]} queryEmbedding - The query embedding vector.
     * @param {NodeWithEmbedding[]} nodes - Array of nodes with embeddings.
     * @param {number} topK - Number of top similar nodes to return.
     * @returns {Array<{node: NodeWithEmbedding, similarity: number}>} Top similar nodes with similarity scores.
     */
    findSimilarNodes(queryEmbedding: number[], nodes: NodeWithEmbedding[], topK?: number): Array<{
        node: NodeWithEmbedding;
        similarity: number;
    }>;
    /**
     * Gets statistics about the embedded nodes by type.
     * @param {NodeWithEmbedding[]} nodes - Array of nodes with embeddings.
     * @returns {Record<string, number>} Node type statistics.
     */
    private getNodeTypeStats;
    /**
     * Generates an embedding for a search query.
     * @param {string} query - The search query text.
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    generateQueryEmbedding(query: string): Promise<number[]>;
    /**
     * Generates an embedding for document content.
     * @param {string} text - The document text.
     * @returns {Promise<number[]>} The generated embedding vector.
     */
    generateDocumentEmbedding(text: string): Promise<number[]>;
    /**
     * Gets embedding service statistics.
     * @returns {Promise<{modelLoaded: boolean, model: string, dimensions: number}>}
     */
    getStats(): Promise<{
        modelLoaded: boolean;
        model: string;
        dimensions: number;
    }>;
}
//# sourceMappingURL=embedding-service.d.ts.map