remcode/dist/vectorizers/embedders/manager.d.ts

Turn your AI assistant into a codebase expert. Intelligent code analysis, semantic search, and software engineering guidance through MCP integration.

import { CodeChunk, EmbeddingManagerOptions } from '../types';
interface ModelInfo {
    id: string;
    name: string;
    embeddingDimension: number;
    strategy: 'code' | 'text';
    apiType: 'feature_extraction' | 'sentence_similarity';
}
export declare class EmbeddingManager {
    private options;
    private hfClient;
    private apiBaseUrl;
    private healthCheckedModels;
    constructor(options: EmbeddingManagerOptions);
    /**
     * Initialize and validate the embedding model
     * Tests the primary model and falls back to alternatives if needed
     * @returns The initialized model ID and configuration
     */
    initializeModel(): Promise<{
        modelId: string;
        modelInfo: ModelInfo;
        isHealthy: boolean;
    }>;
    /**
     * Check if a model is healthy and available via Inference API
     * @param modelId The model ID to check
     * @returns True if the model is available and working
     */
    checkModelHealth(modelId: string): Promise<boolean>;
    /**
     * Get available models with their health status
     * @returns Array of available models with health information
     */
    getAvailableModelsWithHealth(): Promise<Array<ModelInfo & {
        isHealthy: boolean;
    }>>;
    /**
     * Embeds code chunks using the specified model
     * @param chunks Array of code chunks to embed
     * @returns The chunks with embeddings added
     */
    embedChunks(chunks: CodeChunk[]): Promise<CodeChunk[]>;
    /**
     * Embeds a single code chunk
     */
    private embedSingleChunk;
    /**
     * Embeds a single chunk with fallback strategy
     */
    private embedSingleChunkWithFallback;
    /**
     * Gets an embedding from the HuggingFace model
     */
    private getEmbeddingFromModel;
    /**
     * Get embedding via direct API call
     */
    private getEmbeddingViaDirectAPI;
    /**
     * Process embedding result from API based on model type
     */
    private processEmbeddingResult;
    /**
     * Preprocess text for better embedding quality
     */
    private preprocessText;
    /**
     * Get dimension for a specific model
     */
    private getDimensionForModel;
    /**
     * Averages token embeddings to get a single vector
     */
    private averageEmbeddings;
    /**
     * Generates random embeddings as a fallback
     */
    private generateRandomEmbeddings;
    /**
     * Get model information
     */
    getModelInfo(modelId?: string): ModelInfo;
    /**
     * List available models
     */
    getAvailableModels(): ModelInfo[];
}
export {};