@ever_cheng/memory-task-mcp/dist/src/chunking.d.ts

Memory and task management MCP Server

/**
 * Chunking Service for MemTask
 *
 * Intelligent text chunking for better semantic search and embedding generation.
 * Supports multiple chunking strategies optimized for multilingual-e5-large model.
 */
import { Memory } from './types';
/**
 * Memory Chunk Interface
 */
export interface MemoryChunk {
    id: string;
    memoryId: string;
    chunkIndex: number;
    totalChunks: number;
    content: string;
    context: string;
    metadata: {
        created_at: string;
        tags: string[];
        originalLength: number;
        chunkLength: number;
    };
    embedding?: number[];
}
/**
 * Chunking Configuration
 */
export interface ChunkingConfig {
    maxChunkSize: number;
    overlapSize: number;
    minChunkSize: number;
    preserveBoundaries: boolean;
}
/**
 * Default chunking configuration optimized for multilingual-e5-large
 */
export declare const DEFAULT_CHUNKING_CONFIG: ChunkingConfig;
/**
 * Chunking Service Class
 */
export declare class ChunkingService {
    private config;
    constructor(config?: ChunkingConfig);
    /**
     * 智能分塊 - 為 Memory 創建 chunks
     */
    chunkMemory(memory: Memory): Promise<MemoryChunk[]>;
    /**
     * 語義邊界分塊 - 保持段落和句子完整性
     */
    private semanticChunking;
    /**
     * 簡單固定長度分塊
     */
    private simpleChunking;
    /**
     * 將文本分割成句子
     */
    private splitIntoSentences;
    /**
     * 過濾太小的 chunks
     */
    private filterSmallChunks;
    /**
     * 建構上下文信息
     */
    private buildContext;
    /**
     * 獲取前一個 chunk 的上下文
     */
    private getPreviousContext;
    /**
     * 獲取後一個 chunk 的上下文
     */
    private getNextContext;
    /**
     * 批量處理多個 memories
     */
    batchChunkMemories(memories: Memory[]): Promise<MemoryChunk[]>;
    /**
     * 更新配置
     */
    updateConfig(newConfig: Partial<ChunkingConfig>): void;
    /**
     * 獲取當前配置
     */
    getConfig(): ChunkingConfig;
    /**
     * 統計信息
     */
    getChunkingStats(chunks: MemoryChunk[]): {
        totalChunks: number;
        avgChunkLength: number;
        maxChunkLength: number;
        minChunkLength: number;
        totalOriginalLength: number;
        compressionRatio: number;
    };
}