@sylphx/flow 1.1.1 → 1.2.0

package/src/services/search/embeddings-provider.ts

@@ -1,455 +0,0 @@
-/**
- * Embeddings Provider Service - 嵌入向量提供者服務
- * 統一管理所有embedding相關操作
- */
-
-import { type Result, tryCatchAsync } from '../../core/functional/result.js';
-import {
-  EmbeddingInitError,
-  EmbeddingNotInitializedError,
-  type EmbeddingsErrorType,
-  VectorDimensionError,
-} from '../../errors/embeddings-errors.js';
-import type { EmbeddingProvider } from '../../utils/embeddings.js';
-import { getDefaultEmbeddingProvider } from '../../utils/embeddings.js';
-import { chunk } from '../../utils/functional/array.js';
-import { createLogger } from '../../utils/debug-logger.js';
-
-const log = createLogger('search:embeddings');
-
-export interface EmbeddingConfig {
-  provider?: 'openai' | 'local' | 'auto';
-  model?: string;
-  dimensions?: number;
-  batchSize?: number;
-  maxTokens?: number;
-}
-
-export interface EmbeddingResult {
-  embeddings: number[][];
-  tokensUsed: number;
-  processingTime: number;
-}
-
-/**
- * Internal state for embeddings provider
- */
-interface EmbeddingsProviderState {
-  readonly provider?: EmbeddingProvider;
-  readonly isInitialized: boolean;
-}
-
-/**
- * Embeddings Provider Service Interface
- * 嵌入向量提供者服務
- */
-export interface EmbeddingsProviderService {
-  readonly initialize: () => Promise<Result<void, EmbeddingsErrorType>>;
-  readonly generateEmbeddings: (
-    texts: string[]
-  ) => Promise<Result<number[][], EmbeddingsErrorType>>;
-  readonly generateEmbedding: (text: string) => Promise<Result<number[], EmbeddingsErrorType>>;
-  readonly generateEmbeddingsWithMetrics: (
-    texts: string[]
-  ) => Promise<Result<EmbeddingResult, EmbeddingsErrorType>>;
-  readonly isValidText: (text: string) => boolean;
-  readonly preprocessText: (text: string) => string;
-  readonly preprocessTexts: (texts: string[]) => string[];
-  readonly cosineSimilarity: (
-    vecA: number[],
-    vecB: number[]
-  ) => Result<number, VectorDimensionError>;
-  readonly findMostSimilar: (
-    queryVector: number[],
-    candidateVectors: number[][]
-  ) => Result<{ index: number; similarity: number }, VectorDimensionError>;
-  readonly getConfig: () => EmbeddingConfig;
-  readonly updateConfig: (newConfig: Partial<EmbeddingConfig>) => void;
-  readonly getProviderInfo: () => Promise<
-    Result<
-      {
-        provider: string;
-        model: string;
-        dimensions: number;
-        available: boolean;
-      },
-      EmbeddingsErrorType
-    >
-  >;
-}
-
-/**
- * Create embeddings provider service (Factory Function)
- * 提供統一嘅embedding interface和配置管理
- */
-export const createEmbeddingsProviderService = (
-  config: EmbeddingConfig = {}
-): EmbeddingsProviderService => {
-  // Immutable config with defaults
-  let serviceConfig: EmbeddingConfig = {
-    provider: 'auto',
-    batchSize: 10,
-    maxTokens: 8000,
-    dimensions: 1536,
-    ...config,
-  };
-
-  // Mutable state in closure (will be updated immutably)
-  let state: EmbeddingsProviderState = {
-    provider: undefined,
-    isInitialized: false,
-  };
-
-  // Helper: Update state immutably
-  const updateState = (updates: Partial<EmbeddingsProviderState>): void => {
-    state = { ...state, ...updates };
-  };
-
-  /**
-   * 初始化embedding provider
-   */
-  const initialize = async (): Promise<Result<void, EmbeddingsErrorType>> => {
-    if (state.isInitialized) {
-      return { _tag: 'Success', value: undefined };
-    }
-
-    return await tryCatchAsync(
-      async () => {
-        const provider = await getDefaultEmbeddingProvider();
-        updateState({ provider, isInitialized: true });
-        return undefined;
-      },
-      (error) => new EmbeddingInitError(error)
-    );
-  };
-
-  /**
-   * 確保已初始化
-   */
-  const ensureInitialized = async (): Promise<Result<void, EmbeddingsErrorType>> => {
-    if (!state.isInitialized) {
-      return await initialize();
-    }
-    return { _tag: 'Success', value: undefined };
-  };
-
-  /**
-   * 生成嵌入向量
-   */
-  const generateEmbeddings = async (
-    texts: string[]
-  ): Promise<Result<number[][], EmbeddingsErrorType>> => {
-    return await tryCatchAsync(
-      async () => {
-        const initResult = await ensureInitialized();
-        if (initResult._tag === 'Failure') {
-          throw initResult.error;
-        }
-
-        if (!state.provider) {
-          throw new EmbeddingNotInitializedError();
-        }
-
-        // 檢查文本長度
-        const filteredTexts = texts.filter((text) => text.trim().length > 0);
-        if (filteredTexts.length === 0) {
-          return [];
-        }
-
-        // FUNCTIONAL: 批量處理 - Use chunk and flatMap instead of for loop
-        const batchSize = serviceConfig.batchSize || 10;
-        const batches = chunk(batchSize)(filteredTexts);
-
-        // Process batches and flatten results
-        const batchResults = await Promise.all(
-          batches.map(async (batch, index) => {
-            try {
-              return await state.provider?.generateEmbeddings(batch);
-            } catch (error) {
-              log(
-                `Batch ${index * batchSize}-${index * batchSize + batchSize} failed:`,
-                error instanceof Error ? error.message : String(error)
-              );
-              // 為失敗嘅batch添加零向量
-              const zeroVector = new Array(serviceConfig.dimensions || 1536).fill(0);
-              return Array(batch.length).fill(zeroVector);
-            }
-          })
-        );
-
-        return batchResults.flat();
-      },
-      (error) => {
-        if (error instanceof EmbeddingInitError || error instanceof EmbeddingNotInitializedError) {
-          return error;
-        }
-        return new EmbeddingInitError(error);
-      }
-    );
-  };
-
-  /**
-   * 生成單個文本嘅嵌入向量
-   */
-  const generateEmbedding = async (
-    text: string
-  ): Promise<Result<number[], EmbeddingsErrorType>> => {
-    return await tryCatchAsync(
-      async () => {
-        const result = await generateEmbeddings([text]);
-        if (result._tag === 'Failure') {
-          throw result.error;
-        }
-        return result.value[0] || [];
-      },
-      (error) => {
-        if (error instanceof EmbeddingInitError || error instanceof EmbeddingNotInitializedError) {
-          return error;
-        }
-        return new EmbeddingInitError(error);
-      }
-    );
-  };
-
-  /**
-   * 批量生成嵌入向量（帶性能監控）
-   */
-  const generateEmbeddingsWithMetrics = async (
-    texts: string[]
-  ): Promise<Result<EmbeddingResult, EmbeddingsErrorType>> => {
-    return await tryCatchAsync(
-      async () => {
-        const startTime = Date.now();
-
-        const result = await generateEmbeddings(texts);
-        if (result._tag === 'Failure') {
-          throw result.error;
-        }
-
-        const processingTime = Date.now() - startTime;
-        const tokensUsed = estimateTokens(texts);
-
-        return {
-          embeddings: result.value,
-          tokensUsed,
-          processingTime,
-        };
-      },
-      (error) => {
-        if (error instanceof EmbeddingInitError || error instanceof EmbeddingNotInitializedError) {
-          return error;
-        }
-        return new EmbeddingInitError(error);
-      }
-    );
-  };
-
-  /**
-   * 估算token數量
-   */
-  const estimateTokens = (texts: string[]): number => {
-    // 簡單嘅token估算（約4 characters = 1 token）
-    const totalChars = texts.reduce((sum, text) => sum + text.length, 0);
-    return Math.ceil(totalChars / 4);
-  };
-
-  /**
-   * 檢查文本是否適合生成embedding
-   */
-  const isValidText = (text: string): boolean => {
-    if (!text || typeof text !== 'string') {
-      return false;
-    }
-
-    const trimmed = text.trim();
-    if (trimmed.length === 0) {
-      return false;
-    }
-
-    // 檢查長度限制
-    const maxTokens = serviceConfig.maxTokens || 8000;
-    const estimatedTokens = Math.ceil(trimmed.length / 4);
-
-    return estimatedTokens <= maxTokens;
-  };
-
-  /**
-   * 預處理文本
-   */
-  const preprocessText = (text: string): string => {
-    if (!text || typeof text !== 'string') {
-      return '';
-    }
-
-    return (
-      text
-        .trim()
-        // 移除多餘嘅空白字符
-        .replace(/\s+/g, ' ')
-        // 截斷過長嘅文本
-        .slice(0, (serviceConfig.maxTokens || 8000) * 4)
-    );
-  };
-
-  /**
-   * 批量預處理文本
-   */
-  const preprocessTexts = (texts: string[]): string[] => {
-    return texts.filter((text) => isValidText(text)).map((text) => preprocessText(text));
-  };
-
-  /**
-   * 計算兩個向量嘅餘弦相似度
-   */
-  const cosineSimilarity = (
-    vecA: number[],
-    vecB: number[]
-  ): Result<number, VectorDimensionError> => {
-    if (vecA.length !== vecB.length) {
-      return { _tag: 'Failure', error: new VectorDimensionError(vecA.length, vecB.length) };
-    }
-
-    // FUNCTIONAL: Use reduce instead of for loop
-    const { dotProduct, normA, normB } = vecA.reduce(
-      (acc, aVal, i) => {
-        const bVal = vecB[i];
-        return {
-          dotProduct: acc.dotProduct + aVal * bVal,
-          normA: acc.normA + aVal * aVal,
-          normB: acc.normB + bVal * bVal,
-        };
-      },
-      { dotProduct: 0, normA: 0, normB: 0 }
-    );
-
-    if (normA === 0 || normB === 0) {
-      return { _tag: 'Success', value: 0 };
-    }
-
-    return { _tag: 'Success', value: dotProduct / (Math.sqrt(normA) * Math.sqrt(normB)) };
-  };
-
-  /**
-   * 找到最相似嘅向量
-   */
-  const findMostSimilar = (
-    queryVector: number[],
-    candidateVectors: number[][]
-  ): Result<{ index: number; similarity: number }, VectorDimensionError> => {
-    // FUNCTIONAL: Use reduce instead of for loop
-    const result = candidateVectors.reduce<
-      Result<{ index: number; similarity: number }, VectorDimensionError>
-    >(
-      (acc, candidateVector, index) => {
-        // Short-circuit on error
-        if (acc._tag === 'Failure') {
-          return acc;
-        }
-
-        const similarityResult = cosineSimilarity(queryVector, candidateVector);
-        if (similarityResult._tag === 'Failure') {
-          return similarityResult;
-        }
-
-        const similarity = similarityResult.value;
-        return similarity > acc.value.similarity
-          ? { _tag: 'Success', value: { index, similarity } }
-          : acc;
-      },
-      { _tag: 'Success', value: { index: -1, similarity: 0 } }
-    );
-
-    return result;
-  };
-
-  /**
-   * 獲取配置信息
-   */
-  const getConfig = (): EmbeddingConfig => {
-    return { ...serviceConfig };
-  };
-
-  /**
-   * 更新配置
-   */
-  const updateConfig = (newConfig: Partial<EmbeddingConfig>): void => {
-    serviceConfig = { ...serviceConfig, ...newConfig };
-
-    // 如果provider相關配置改變，需要重新初始化
-    if (newConfig.provider || newConfig.model) {
-      updateState({ isInitialized: false });
-    }
-  };
-
-  /**
-   * 獲取provider信息
-   */
-  const getProviderInfo = async (): Promise<
-    Result<
-      {
-        provider: string;
-        model: string;
-        dimensions: number;
-        available: boolean;
-      },
-      EmbeddingsErrorType
-    >
-  > => {
-    return await tryCatchAsync(
-      async () => {
-        const initResult = await ensureInitialized();
-        if (initResult._tag === 'Failure') {
-          // Return unavailable info instead of throwing
-          return {
-            provider: serviceConfig.provider || 'unknown',
-            model: serviceConfig.model || 'unknown',
-            dimensions: serviceConfig.dimensions || 1536,
-            available: false,
-          };
-        }
-
-        if (!state.provider) {
-          return {
-            provider: serviceConfig.provider || 'unknown',
-            model: serviceConfig.model || 'unknown',
-            dimensions: serviceConfig.dimensions || 1536,
-            available: false,
-          };
-        }
-
-        // 這裡需要根據實際provider interface實現
-        return {
-          provider: serviceConfig.provider || 'unknown',
-          model: serviceConfig.model || 'unknown',
-          dimensions: serviceConfig.dimensions || 1536,
-          available: true,
-        };
-      },
-      (error) => new EmbeddingInitError(error)
-    );
-  };
-
-  // Return service interface
-  return {
-    initialize,
-    generateEmbeddings,
-    generateEmbedding,
-    generateEmbeddingsWithMetrics,
-    isValidText,
-    preprocessText,
-    preprocessTexts,
-    cosineSimilarity,
-    findMostSimilar,
-    getConfig,
-    updateConfig,
-    getProviderInfo,
-  };
-};
-
-// 預設實例 - Use factory function
-export const defaultEmbeddingsProvider = createEmbeddingsProviderService();
-
-// 工廠函數別名 - For backwards compatibility
-export function createEmbeddingsProvider(config?: EmbeddingConfig): EmbeddingsProviderService {
-  return createEmbeddingsProviderService(config);
-}