memento-mcp-server 1.6.0 → 1.7.2

package/src/services/embedding-service.spec.ts

@@ -1,342 +0,0 @@
-import { describe, it, expect, beforeEach, vi, afterEach } from 'vitest';
-import { EmbeddingService } from './embedding-service.js';
-import { mementoConfig } from '../config/index.js';
-
-// Mock dependencies
-vi.mock('../config/index.js', () => ({
-  mementoConfig: {
-    openaiApiKey: 'test-key',
-    embeddingProvider: 'openai'
-  }
-}));
-
-vi.mock('./gemini-embedding-service.js', () => ({
-  GeminiEmbeddingService: vi.fn().mockImplementation(() => ({
-    generateEmbedding: vi.fn().mockResolvedValue({
-      embedding: [0.1, 0.2, 0.3],
-      model: 'gemini-model',
-      usage: { prompt_tokens: 10, total_tokens: 10 }
-    }),
-    isAvailable: vi.fn().mockReturnValue(true),
-    getModelInfo: vi.fn().mockReturnValue({
-      model: 'gemini-model',
-      dimensions: 768,
-      maxTokens: 1000
-    })
-  }))
-}));
-
-vi.mock('./lightweight-embedding-service.js', () => ({
-  LightweightEmbeddingService: vi.fn().mockImplementation(() => ({
-    generateEmbedding: vi.fn().mockResolvedValue({
-      embedding: [0.4, 0.5, 0.6],
-      model: 'lightweight-model',
-      usage: { prompt_tokens: 5, total_tokens: 5 }
-    }),
-    isAvailable: vi.fn().mockReturnValue(true),
-    getModelInfo: vi.fn().mockReturnValue({
-      model: 'lightweight-model',
-      dimensions: 512,
-      maxTokens: 2000
-    })
-  }))
-}));
-
-// Mock OpenAI
-const mockOpenAI = {
-  embeddings: {
-    create: vi.fn()
-  }
-};
-
-vi.mock('openai', () => ({
-  default: vi.fn().mockImplementation(() => mockOpenAI)
-}));
-
-describe('EmbeddingService', () => {
-  let embeddingService: EmbeddingService;
-
-  beforeEach(() => {
-    vi.clearAllMocks();
-    embeddingService = new EmbeddingService();
-  });
-
-  afterEach(() => {
-    vi.restoreAllMocks();
-  });
-
-  describe('생성자', () => {
-    it('서비스가 올바르게 초기화되어야 함', () => {
-      expect(embeddingService).toBeInstanceOf(EmbeddingService);
-    });
-
-    it('OpenAI API 키가 없으면 경고를 출력해야 함', () => {
-      const consoleSpy = vi.spyOn(console, 'warn');
-      vi.mocked(mementoConfig).openaiApiKey = undefined;
-      
-      new EmbeddingService();
-      
-      expect(consoleSpy).toHaveBeenCalledWith('⚠️ OpenAI API 키가 설정되지 않았습니다. 임베딩 기능이 비활성화됩니다.');
-    });
-  });
-
-  describe('generateEmbedding', () => {
-    it('빈 텍스트에 대해 에러를 던져야 함', async () => {
-      await expect(embeddingService.generateEmbedding('')).rejects.toThrow('텍스트가 비어있습니다');
-      await expect(embeddingService.generateEmbedding('   ')).rejects.toThrow('텍스트가 비어있습니다');
-    });
-
-    it('OpenAI 임베딩을 성공적으로 생성해야 함', async () => {
-      const mockResponse = {
-        data: [{ embedding: [0.1, 0.2, 0.3] }],
-        usage: { prompt_tokens: 10, total_tokens: 10 }
-      };
-      
-      mockOpenAI.embeddings.create.mockResolvedValue(mockResponse);
-
-      const result = await embeddingService.generateEmbedding('test text');
-
-      expect(result).toEqual({
-        embedding: [0.1, 0.2, 0.3],
-        model: 'text-embedding-3-small',
-        usage: { prompt_tokens: 10, total_tokens: 10 }
-      });
-    });
-
-    it('Gemini 임베딩을 성공적으로 생성해야 함', async () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'gemini';
-
-      const result = await embeddingService.generateEmbedding('test text');
-
-      expect(result).toEqual({
-        embedding: [0.1, 0.2, 0.3],
-        model: 'gemini-model',
-        usage: { prompt_tokens: 10, total_tokens: 10 }
-      });
-    });
-
-    it('경량 임베딩을 성공적으로 생성해야 함', async () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'lightweight';
-
-      const result = await embeddingService.generateEmbedding('test text');
-
-      expect(result).toEqual({
-        embedding: [0.4, 0.5, 0.6],
-        model: 'lightweight-model',
-        usage: { prompt_tokens: 5, total_tokens: 5 }
-      });
-    });
-
-    it('알 수 없는 제공자에 대해 경량 서비스로 fallback해야 함', async () => {
-      const consoleSpy = vi.spyOn(console, 'warn');
-      vi.mocked(mementoConfig).embeddingProvider = 'unknown' as any;
-
-      const result = await embeddingService.generateEmbedding('test text');
-
-      expect(consoleSpy).toHaveBeenCalledWith('⚠️ 알 수 없는 임베딩 제공자: unknown, 경량 서비스로 fallback');
-      expect(result).toEqual({
-        embedding: [0.4, 0.5, 0.6],
-        model: 'lightweight-model',
-        usage: { prompt_tokens: 5, total_tokens: 5 }
-      });
-    });
-
-    it('OpenAI 실패 시 경량 서비스로 fallback해야 함', async () => {
-      const consoleSpy = vi.spyOn(console, 'warn');
-      mockOpenAI.embeddings.create.mockRejectedValue(new Error('OpenAI API error'));
-
-      const result = await embeddingService.generateEmbedding('test text');
-
-      expect(consoleSpy).toHaveBeenCalledWith('⚠️ openai 임베딩 실패, 경량 서비스로 fallback:', expect.any(Error));
-      expect(result).toEqual({
-        embedding: [0.4, 0.5, 0.6],
-        model: 'lightweight-model',
-        usage: { prompt_tokens: 5, total_tokens: 5 }
-      });
-    });
-
-    it('캐시에서 결과를 반환해야 함', async () => {
-      const mockResponse = {
-        data: [{ embedding: [0.1, 0.2, 0.3] }],
-        usage: { prompt_tokens: 10, total_tokens: 10 }
-      };
-      
-      mockOpenAI.embeddings.create.mockResolvedValue(mockResponse);
-
-      // 첫 번째 호출
-      const result1 = await embeddingService.generateEmbedding('test text');
-      // 두 번째 호출 (캐시에서 반환)
-      const result2 = await embeddingService.generateEmbedding('test text');
-
-      expect(mockOpenAI.embeddings.create).toHaveBeenCalledTimes(1);
-      expect(result1).toEqual(result2);
-    });
-  });
-
-  describe('searchSimilar', () => {
-    beforeEach(async () => {
-      const mockResponse = {
-        data: [{ embedding: [0.1, 0.2, 0.3] }],
-        usage: { prompt_tokens: 10, total_tokens: 10 }
-      };
-      mockOpenAI.embeddings.create.mockResolvedValue(mockResponse);
-    });
-
-    it('유사한 임베딩을 검색해야 함', async () => {
-      const embeddings = [
-        { id: '1', content: 'test content 1', embedding: [0.1, 0.2, 0.3] },
-        { id: '2', content: 'test content 2', embedding: [0.9, 0.8, 0.7] },
-        { id: '3', content: 'test content 3', embedding: [0.2, 0.3, 0.4] }
-      ];
-
-      const results = await embeddingService.searchSimilar('test query', embeddings, 2, 0.5);
-
-      expect(results).toHaveLength(2);
-      expect(results[0].similarity).toBeGreaterThanOrEqual(results[1].similarity);
-    });
-
-    it('임계값보다 낮은 유사도는 필터링해야 함', async () => {
-      const embeddings = [
-        { id: '1', content: 'test content 1', embedding: [0.1, 0.2, 0.3] },
-        { id: '2', content: 'test content 2', embedding: [0.9, 0.8, 0.7] }
-      ];
-
-      const results = await embeddingService.searchSimilar('test query', embeddings, 10, 0.9);
-
-      expect(results).toHaveLength(0);
-    });
-
-    it('쿼리 임베딩 생성 실패 시 빈 배열을 반환해야 함', async () => {
-      mockOpenAI.embeddings.create.mockRejectedValue(new Error('API error'));
-
-      const embeddings = [
-        { id: '1', content: 'test content 1', embedding: [0.1, 0.2, 0.3] }
-      ];
-
-      const results = await embeddingService.searchSimilar('test query', embeddings);
-
-      expect(results).toEqual([]);
-    });
-  });
-
-  describe('cosineSimilarity', () => {
-    it('동일한 벡터의 유사도는 1이어야 함', () => {
-      const vector = [1, 2, 3];
-      const result = (embeddingService as any).cosineSimilarity(vector, vector);
-      expect(result).toBeCloseTo(1, 5);
-    });
-
-    it('직교하는 벡터의 유사도는 0이어야 함', () => {
-      const vector1 = [1, 0, 0];
-      const vector2 = [0, 1, 0];
-      const result = (embeddingService as any).cosineSimilarity(vector1, vector2);
-      expect(result).toBeCloseTo(0, 5);
-    });
-
-    it('다른 차원의 벡터에 대해 에러를 던져야 함', () => {
-      const vector1 = [1, 2];
-      const vector2 = [1, 2, 3];
-      
-      expect(() => {
-        (embeddingService as any).cosineSimilarity(vector1, vector2);
-      }).toThrow('벡터 차원이 일치하지 않습니다');
-    });
-
-    it('영벡터에 대해 0을 반환해야 함', () => {
-      const vector1 = [0, 0, 0];
-      const vector2 = [1, 2, 3];
-      const result = (embeddingService as any).cosineSimilarity(vector1, vector2);
-      expect(result).toBe(0);
-    });
-  });
-
-  describe('isAvailable', () => {
-    it('OpenAI 제공자가 사용 가능해야 함', () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'openai';
-      expect(embeddingService.isAvailable()).toBe(true);
-    });
-
-    it('Gemini 제공자가 사용 가능해야 함', () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'gemini';
-      expect(embeddingService.isAvailable()).toBe(true);
-    });
-
-    it('경량 제공자가 사용 가능해야 함', () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'lightweight';
-      expect(embeddingService.isAvailable()).toBe(true);
-    });
-  });
-
-  describe('getModelInfo', () => {
-    it('OpenAI 모델 정보를 반환해야 함', () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'openai';
-      
-      const info = embeddingService.getModelInfo();
-      
-      expect(info).toEqual({
-        model: 'text-embedding-3-small',
-        dimensions: 1536,
-        maxTokens: 8191
-      });
-    });
-
-    it('Gemini 모델 정보를 반환해야 함', () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'gemini';
-      
-      const info = embeddingService.getModelInfo();
-      
-      expect(info).toEqual({
-        model: 'gemini-model',
-        dimensions: 768,
-        maxTokens: 1000
-      });
-    });
-
-    it('경량 모델 정보를 반환해야 함', () => {
-      vi.mocked(mementoConfig).embeddingProvider = 'lightweight';
-      
-      const info = embeddingService.getModelInfo();
-      
-      expect(info).toEqual({
-        model: 'lightweight-model',
-        dimensions: 512,
-        maxTokens: 2000
-      });
-    });
-  });
-
-  describe('캐시 관리', () => {
-    it('캐시 키를 올바르게 생성해야 함', () => {
-      const key1 = (embeddingService as any).generateCacheKey('test text');
-      const key2 = (embeddingService as any).generateCacheKey('test text');
-      const key3 = (embeddingService as any).generateCacheKey('different text');
-      
-      expect(key1).toBe(key2);
-      expect(key1).not.toBe(key3);
-    });
-
-    it('텍스트 해시를 올바르게 생성해야 함', () => {
-      const hash1 = (embeddingService as any).hashText('test');
-      const hash2 = (embeddingService as any).hashText('test');
-      const hash3 = (embeddingService as any).hashText('different');
-      
-      expect(hash1).toBe(hash2);
-      expect(hash1).not.toBe(hash3);
-      expect(typeof hash1).toBe('string');
-    });
-  });
-
-  describe('텍스트 자르기', () => {
-    it('짧은 텍스트는 그대로 반환해야 함', () => {
-      const shortText = 'short text';
-      const result = (embeddingService as any).truncateText(shortText);
-      expect(result).toBe(shortText);
-    });
-
-    it('긴 텍스트는 잘라야 함', () => {
-      const longText = 'a'.repeat(40000); // 8191 * 4보다 긴 텍스트
-      const result = (embeddingService as any).truncateText(longText);
-      expect(result.length).toBeLessThanOrEqual(8191 * 4);
-    });
-  });
-});

package/src/services/embedding-service.ts

@@ -1,333 +0,0 @@
-/**
- * 통합 임베딩 서비스 (레거시)
- * 새로운 UnifiedEmbeddingService로 대체됨
- * 
- * @deprecated 이 클래스는 더 이상 사용되지 않습니다.
- * @see UnifiedEmbeddingService 새로운 통합 서비스를 사용하세요.
- * @since 0.1.0
- * @removed 0.2.0 이 클래스는 제거될 예정입니다.
- */
-
-import OpenAI from 'openai';
-import { mementoConfig } from '../config/index.js';
-import { GeminiEmbeddingService, type GeminiEmbeddingResult, type GeminiSimilarityResult } from './gemini-embedding-service.js';
-import { LightweightEmbeddingService, type LightweightEmbeddingResult, type LightweightSimilarityResult } from './lightweight-embedding-service.js';
-
-export interface EmbeddingResult {
-  embedding: number[];
-  model: string;
-  usage: {
-    prompt_tokens: number;
-    total_tokens: number;
-  };
-}
-
-export interface SimilarityResult {
-  id: string;
-  content: string;
-  similarity: number;
-  score: number;
-}
-
-export class EmbeddingService {
-  private openai: OpenAI | null = null;
-  private geminiService: GeminiEmbeddingService;
-  private lightweightService: LightweightEmbeddingService;
-  private readonly model = 'text-embedding-3-small'; // 1536차원
-  private readonly maxTokens = 8191; // text-embedding-3-small 최대 토큰
-  private embeddingCache: Map<string, EmbeddingResult> = new Map(); // 임베딩 캐시
-  private batchQueue: string[] = []; // 배치 처리 큐
-  private batchTimeout: NodeJS.Timeout | null = null;
-
-  constructor() {
-    this.geminiService = new GeminiEmbeddingService();
-    this.lightweightService = new LightweightEmbeddingService();
-    this.initializeOpenAI();
-  }
-
-  /**
-   * OpenAI 클라이언트 초기화
-   */
-  private initializeOpenAI(): void {
-    if (!mementoConfig.openaiApiKey) {
-      console.warn('⚠️ OpenAI API 키가 설정되지 않았습니다. 임베딩 기능이 비활성화됩니다.');
-      return;
-    }
-
-    try {
-      this.openai = new OpenAI({
-        apiKey: mementoConfig.openaiApiKey,
-      });
-      console.log('✅ OpenAI 임베딩 서비스 초기화 완료');
-    } catch (error) {
-      console.error('❌ OpenAI 초기화 실패:', error);
-      this.openai = null;
-    }
-  }
-
-  /**
-   * 텍스트를 임베딩으로 변환 - 캐시 최적화
-   */
-  async generateEmbedding(text: string): Promise<EmbeddingResult | null> {
-    if (!text || text.trim().length === 0) {
-      throw new Error('텍스트가 비어있습니다');
-    }
-
-    // 1. 캐시 확인
-    const cacheKey = this.generateCacheKey(text);
-    const cached = this.embeddingCache.get(cacheKey);
-    if (cached) {
-      return cached;
-    }
-
-    // 2. 설정에 따른 제공자 선택
-    const provider = mementoConfig.embeddingProvider;
-    let result: EmbeddingResult | null = null;
-
-    try {
-      switch (provider) {
-        case 'openai':
-          result = await this.generateOpenAIEmbedding(text);
-          break;
-        case 'gemini':
-          result = await this.generateGeminiEmbedding(text);
-          break;
-        case 'lightweight':
-          result = await this.generateLightweightEmbedding(text);
-          break;
-        default:
-          console.warn(`⚠️ 알 수 없는 임베딩 제공자: ${provider}, 경량 서비스로 fallback`);
-          result = await this.generateLightweightEmbedding(text);
-      }
-    } catch (error) {
-      console.warn(`⚠️ ${provider} 임베딩 실패, 경량 서비스로 fallback:`, error);
-      result = await this.generateLightweightEmbedding(text);
-    }
-
-    if (result) {
-      // 캐시에 저장
-      this.embeddingCache.set(cacheKey, result);
-      this.cleanupCache();
-    }
-
-    return result;
-  }
-
-  /**
-   * OpenAI 임베딩 생성
-   */
-  private async generateOpenAIEmbedding(text: string): Promise<EmbeddingResult | null> {
-    if (!this.openai) {
-      throw new Error('OpenAI 클라이언트가 초기화되지 않았습니다');
-    }
-
-    const truncatedText = this.truncateText(text);
-    
-    const response = await this.openai.embeddings.create({
-      model: this.model,
-      input: truncatedText,
-      encoding_format: 'float',
-    });
-
-    const embedding = response.data[0]?.embedding;
-    if (!embedding) {
-      throw new Error('임베딩 생성에 실패했습니다');
-    }
-    
-    return {
-      embedding,
-      model: this.model,
-      usage: {
-        prompt_tokens: response.usage.prompt_tokens,
-        total_tokens: response.usage.total_tokens,
-      },
-    };
-  }
-
-  /**
-   * Gemini 임베딩 생성
-   */
-  private async generateGeminiEmbedding(text: string): Promise<EmbeddingResult | null> {
-    const geminiResult = await this.geminiService.generateEmbedding(text);
-    if (!geminiResult) {
-      return null;
-    }
-
-    return {
-      embedding: geminiResult.embedding,
-      model: geminiResult.model,
-      usage: geminiResult.usage,
-    };
-  }
-
-  /**
-   * 경량 임베딩 생성
-   */
-  private async generateLightweightEmbedding(text: string): Promise<EmbeddingResult | null> {
-    const lightweightResult = await this.lightweightService.generateEmbedding(text);
-    if (!lightweightResult) {
-      return null;
-    }
-
-    return {
-      embedding: lightweightResult.embedding,
-      model: lightweightResult.model,
-      usage: lightweightResult.usage,
-    };
-  }
-
-  /**
-   * 쿼리와 유사한 임베딩 검색
-   */
-  async searchSimilar(
-    query: string,
-    embeddings: Array<{ id: string; content: string; embedding: number[] }>,
-    limit: number = 10,
-    threshold: number = 0.7
-  ): Promise<SimilarityResult[]> {
-    // 쿼리 임베딩 생성 (fallback 로직 포함)
-    const queryEmbedding = await this.generateEmbedding(query);
-    if (!queryEmbedding) {
-      return [];
-    }
-
-    // 코사인 유사도 계산
-    const similarities = embeddings.map(item => {
-      const similarity = this.cosineSimilarity(queryEmbedding.embedding, item.embedding);
-      return {
-        id: item.id,
-        content: item.content,
-        similarity,
-        score: similarity
-      };
-    });
-
-    // 유사도 순으로 정렬하고 임계값 필터링
-    return similarities
-      .filter(item => item.similarity >= threshold)
-      .sort((a, b) => b.similarity - a.similarity)
-      .slice(0, limit);
-  }
-
-  /**
-   * 코사인 유사도 계산
-   */
-  private cosineSimilarity(a: number[], b: number[]): number {
-    if (a.length !== b.length) {
-      throw new Error('벡터 차원이 일치하지 않습니다');
-    }
-
-    let dotProduct = 0;
-    let normA = 0;
-    let normB = 0;
-
-    for (let i = 0; i < a.length; i++) {
-      const aVal = a[i] || 0;
-      const bVal = b[i] || 0;
-      dotProduct += aVal * bVal;
-      normA += aVal * aVal;
-      normB += bVal * bVal;
-    }
-
-    normA = Math.sqrt(normA);
-    normB = Math.sqrt(normB);
-
-    if (normA === 0 || normB === 0) {
-      return 0;
-    }
-
-    return dotProduct / (normA * normB);
-  }
-
-  /**
-   * 캐시 키 생성
-   */
-  private generateCacheKey(text: string): string {
-    // 텍스트 해시를 사용하여 캐시 키 생성
-    return `embedding:${this.hashText(text)}`;
-  }
-
-  /**
-   * 텍스트 해시 생성
-   */
-  private hashText(text: string): string {
-    let hash = 0;
-    for (let i = 0; i < text.length; i++) {
-      const char = text.charCodeAt(i);
-      hash = ((hash << 5) - hash) + char;
-      hash = hash & hash; // 32비트 정수로 변환
-    }
-    return Math.abs(hash).toString(36);
-  }
-
-  /**
-   * 캐시 정리
-   */
-  private cleanupCache(): void {
-    const maxCacheSize = 1000;
-    if (this.embeddingCache.size > maxCacheSize) {
-      const entries = Array.from(this.embeddingCache.entries());
-      this.embeddingCache.clear();
-      // 최신 500개만 유지
-      entries.slice(-500).forEach(([key, value]) => {
-        this.embeddingCache.set(key, value);
-      });
-    }
-  }
-
-  /**
-   * 텍스트를 토큰 제한에 맞게 자르기
-   */
-  private truncateText(text: string): string {
-    // 간단한 토큰 추정 (1 토큰 ≈ 4 문자)
-    const estimatedTokens = text.length / 4;
-    
-    if (estimatedTokens <= this.maxTokens) {
-      return text;
-    }
-
-    // 토큰 제한에 맞게 자르기
-    const maxChars = this.maxTokens * 4;
-    return text.substring(0, maxChars);
-  }
-
-  /**
-   * 서비스 사용 가능 여부 확인
-   */
-  isAvailable(): boolean {
-    const provider = mementoConfig.embeddingProvider;
-    
-    switch (provider) {
-      case 'openai':
-        return this.openai !== null;
-      case 'gemini':
-        return this.geminiService.isAvailable();
-      case 'lightweight':
-        return this.lightweightService.isAvailable();
-      default:
-        return this.lightweightService.isAvailable();
-    }
-  }
-
-  /**
-   * 모델 정보 반환
-   */
-  getModelInfo(): { model: string; dimensions: number; maxTokens: number } {
-    const provider = mementoConfig.embeddingProvider;
-    
-    switch (provider) {
-      case 'openai':
-        return {
-          model: this.model,
-          dimensions: 1536, // text-embedding-3-small 차원
-          maxTokens: this.maxTokens,
-        };
-      case 'gemini':
-        return this.geminiService.getModelInfo();
-      case 'lightweight':
-        return this.lightweightService.getModelInfo();
-      default:
-        return this.lightweightService.getModelInfo();
-    }
-  }
-}