/**
 * Local embedding adapter
 *
 * Provides embedding generation without external API calls.
 * Uses a simple but effective TF-IDF based approach for semantic similarity.
 *
 * For production use with better quality, consider using transformers.js
 * with a model like 'Xenova/all-MiniLM-L6-v2'.
 *
 * @packageDocumentation
 */
import type { EmbeddingProvider, LocalEmbeddingConfig, EmbeddingResult, BatchEmbeddingResult } from '../types/graphrag.js';
import { BaseEmbeddingAdapter } from './base.js';
/**
 * Local embedding adapter using TF-IDF-like approach
 *
 * This provides a simple, fast, offline embedding that works well for
 * finding similar skills based on shared vocabulary and structure.
 */
export declare class LocalEmbeddingAdapter extends BaseEmbeddingAdapter {
    private readonly modelName;
    private readonly dims;
    private vocabulary;
    private idfScores;
    private documentCount;
    constructor(config: LocalEmbeddingConfig);
    get provider(): EmbeddingProvider;
    get model(): string;
    get dimensions(): number;
    isConfigured(): boolean;
    /**
     * Tokenize text into normalized tokens
     */
    private tokenize;
    /**
     * Calculate term frequency for a document
     */
    private calculateTF;
    /**
     * Generate a deterministic embedding vector from text
     *
     * Uses a combination of:
     * 1. Token-based hashing for semantic representation
     * 2. N-gram features for capturing word relationships
     * 3. Positional encoding for word order sensitivity
     */
    private generateEmbedding;
    /**
     * Extract character n-grams from text
     */
    private getCharNgrams;
    embed(text: string): Promise<EmbeddingResult>;
    embedBatch(texts: string[]): Promise<BatchEmbeddingResult>;
    /**
     * Build vocabulary from a corpus for improved embeddings
     *
     * Call this with your skill names/descriptions to improve
     * the quality of embeddings through proper IDF weighting.
     */
    buildVocabulary(documents: string[]): void;
    /**
     * Get vocabulary size
     */
    getVocabularySize(): number;
}
/**
 * Create a local embedding adapter
 *
 * @example
 * ```typescript
 * // Create adapter
 * const adapter = createLocalEmbeddingAdapter();
 *
 * // Optionally build vocabulary for better quality
 * adapter.buildVocabulary(skillDescriptions);
 *
 * // Generate embedding
 * const result = await adapter.embed('Learn to write JavaScript functions');
 * ```
 */
export declare function createLocalEmbeddingAdapter(model?: string): LocalEmbeddingAdapter;
//# sourceMappingURL=local.d.ts.map