@langchain/community/dist/embeddings/alibaba_tongyi.cjs.map

Third-party integrations for LangChain.js

{"version":3,"file":"alibaba_tongyi.cjs","names":["Embeddings"],"sources":["../../src/embeddings/alibaba_tongyi.ts"],"sourcesContent":["import { getEnvironmentVariable } from \"@langchain/core/utils/env\";\nimport { Embeddings, type EmbeddingsParams } from \"@langchain/core/embeddings\";\nimport { chunkArray } from \"@langchain/core/utils/chunk_array\";\n\nexport interface AlibabaTongyiEmbeddingsParams extends EmbeddingsParams {\n  /** Model name to use */\n  modelName:\n    | \"multimodal-embedding-v1\"\n    | \"text-embedding-v1\"\n    | \"text-embedding-v2\"\n    | \"text-embedding-v3\"\n    | \"text-embedding-v4\";\n\n  /**\n   * Timeout to use when making requests to AlibabaTongyi.\n   */\n  timeout?: number;\n\n  /**\n   * The maximum number of documents to embed in a single request. This is\n   * limited by the AlibabaTongyi API to a maximum of 2048.\n   */\n  batchSize?: number;\n\n  /**\n   * Whether to strip new lines from the input text.\n   */\n  stripNewLines?: boolean;\n\n  parameters?: {\n    /**\n     * 取值：query 或者 document，默认值为 document\n     * 说明：文本转换为向量后可以应用于检索、聚类、分类等下游任务，\n     * \t对检索这类非对称任务为了达到更好的检索效果建议区分查询文本（query）和\n     * \t底库文本（document）类型, 聚类、分类等对称任务可以不用特殊指定，\n     * \t采用系统默认值\"document\"即可\n     */\n    text_type?: \"query\" | \"document\";\n  };\n}\n\ninterface EmbeddingCreateParams {\n  model: AlibabaTongyiEmbeddingsParams[\"modelName\"];\n  input: {\n    texts: string[];\n  };\n\n  parameters?: AlibabaTongyiEmbeddingsParams[\"parameters\"];\n}\n\ninterface EmbeddingResponse {\n  output: {\n    embeddings: { text_index: number; embedding: number[] }[];\n  };\n\n  usage: {\n    total_tokens: number;\n  };\n\n  request_id: string;\n}\n\ninterface EmbeddingErrorResponse {\n  code: string;\n  message: string;\n  request_id: string;\n}\n\nexport class AlibabaTongyiEmbeddings\n  extends Embeddings\n  implements AlibabaTongyiEmbeddingsParams\n{\n  modelName: AlibabaTongyiEmbeddingsParams[\"modelName\"] = \"text-embedding-v2\";\n\n  batchSize = 24;\n\n  stripNewLines = true;\n\n  apiKey: string;\n\n  parameters: EmbeddingCreateParams[\"parameters\"];\n\n  constructor(\n    fields?: Partial<AlibabaTongyiEmbeddingsParams> & {\n      verbose?: boolean;\n      apiKey?: string;\n    }\n  ) {\n    const fieldsWithDefaults = { maxConcurrency: 2, ...fields };\n    super(fieldsWithDefaults);\n\n    const apiKey =\n      fieldsWithDefaults?.apiKey ?? getEnvironmentVariable(\"ALIBABA_API_KEY\");\n\n    if (!apiKey) throw new Error(\"AlibabaAI API key not found\");\n\n    this.apiKey = apiKey;\n\n    this.modelName = fieldsWithDefaults?.modelName ?? this.modelName;\n    this.batchSize = fieldsWithDefaults?.batchSize ?? this.batchSize;\n    this.stripNewLines =\n      fieldsWithDefaults?.stripNewLines ?? this.stripNewLines;\n\n    this.parameters = {\n      text_type: fieldsWithDefaults?.parameters?.text_type ?? \"document\",\n    };\n  }\n\n  /**\n   * Method to generate embeddings for an array of documents. Splits the\n   * documents into batches and makes requests to the AlibabaTongyi API to generate\n   * embeddings.\n   * @param texts Array of documents to generate embeddings for.\n   * @returns Promise that resolves to a 2D array of embeddings for each document.\n   */\n  async embedDocuments(texts: string[]): Promise<number[][]> {\n    const batches = chunkArray(\n      this.stripNewLines ? texts.map((t) => t.replace(/\\n/g, \" \")) : texts,\n      this.batchSize\n    );\n    const batchRequests = batches.map((batch) => {\n      const params = this.getParams(batch);\n\n      return this.embeddingWithRetry(params);\n    });\n\n    const batchResponses = await Promise.all(batchRequests);\n    const embeddings: number[][] = [];\n\n    for (let i = 0; i < batchResponses.length; i += 1) {\n      const batch = batches[i];\n      const batchResponse = batchResponses[i] || [];\n      for (let j = 0; j < batch.length; j += 1) {\n        embeddings.push(batchResponse[j]);\n      }\n    }\n\n    return embeddings;\n  }\n\n  /**\n   * Method to generate an embedding for a single document. Calls the\n   * embeddingWithRetry method with the document as the input.\n   * @param text Document to generate an embedding for.\n   * @returns Promise that resolves to an embedding for the document.\n   */\n  async embedQuery(text: string): Promise<number[]> {\n    const params = this.getParams([\n      this.stripNewLines ? text.replace(/\\n/g, \" \") : text,\n    ]);\n\n    const embeddings = (await this.embeddingWithRetry(params)) || [[]];\n    return embeddings[0];\n  }\n\n  /**\n   * Method to generate an embedding params.\n   * @param texts Array of documents to generate embeddings for.\n   * @returns an embedding params.\n   */\n  private getParams(\n    texts: EmbeddingCreateParams[\"input\"][\"texts\"]\n  ): EmbeddingCreateParams {\n    return {\n      model: this.modelName,\n      input: {\n        texts,\n      },\n      parameters: this.parameters,\n    };\n  }\n\n  /**\n   * Private method to make a request to the OpenAI API to generate\n   * embeddings. Handles the retry logic and returns the response from the\n   * API.\n   * @param request Request to send to the OpenAI API.\n   * @returns Promise that resolves to the response from the API.\n   */\n  private async embeddingWithRetry(body: EmbeddingCreateParams) {\n    return fetch(\n      \"https://dashscope.aliyuncs.com/api/v1/services/embeddings/text-embedding/text-embedding\",\n      {\n        method: \"POST\",\n        headers: {\n          \"Content-Type\": \"application/json\",\n          Authorization: `Bearer ${this.apiKey}`,\n        },\n        body: JSON.stringify(body),\n      }\n    ).then(async (response) => {\n      const embeddingData: EmbeddingResponse | EmbeddingErrorResponse =\n        await response.json();\n\n      if (\"code\" in embeddingData && embeddingData.code) {\n        throw new Error(`${embeddingData.code}: ${embeddingData.message}`);\n      }\n\n      return (embeddingData as EmbeddingResponse).output.embeddings.map(\n        ({ embedding }) => embedding\n      );\n    });\n  }\n}\n"],"mappings":";;;;;;;AAoEA,IAAa,0BAAb,cACUA,2BAAAA,WAEV;CACE,YAAwD;CAExD,YAAY;CAEZ,gBAAgB;CAEhB;CAEA;CAEA,YACE,QAIA;EACA,MAAM,qBAAqB;GAAE,gBAAgB;GAAG,GAAG;GAAQ;AAC3D,QAAM,mBAAmB;EAEzB,MAAM,SACJ,oBAAoB,WAAA,GAAA,0BAAA,wBAAiC,kBAAkB;AAEzE,MAAI,CAAC,OAAQ,OAAM,IAAI,MAAM,8BAA8B;AAE3D,OAAK,SAAS;AAEd,OAAK,YAAY,oBAAoB,aAAa,KAAK;AACvD,OAAK,YAAY,oBAAoB,aAAa,KAAK;AACvD,OAAK,gBACH,oBAAoB,iBAAiB,KAAK;AAE5C,OAAK,aAAa,EAChB,WAAW,oBAAoB,YAAY,aAAa,YACzD;;;;;;;;;CAUH,MAAM,eAAe,OAAsC;EACzD,MAAM,WAAA,GAAA,kCAAA,YACJ,KAAK,gBAAgB,MAAM,KAAK,MAAM,EAAE,QAAQ,OAAO,IAAI,CAAC,GAAG,OAC/D,KAAK,UACN;EACD,MAAM,gBAAgB,QAAQ,KAAK,UAAU;GAC3C,MAAM,SAAS,KAAK,UAAU,MAAM;AAEpC,UAAO,KAAK,mBAAmB,OAAO;IACtC;EAEF,MAAM,iBAAiB,MAAM,QAAQ,IAAI,cAAc;EACvD,MAAM,aAAyB,EAAE;AAEjC,OAAK,IAAI,IAAI,GAAG,IAAI,eAAe,QAAQ,KAAK,GAAG;GACjD,MAAM,QAAQ,QAAQ;GACtB,MAAM,gBAAgB,eAAe,MAAM,EAAE;AAC7C,QAAK,IAAI,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK,EACrC,YAAW,KAAK,cAAc,GAAG;;AAIrC,SAAO;;;;;;;;CAST,MAAM,WAAW,MAAiC;EAChD,MAAM,SAAS,KAAK,UAAU,CAC5B,KAAK,gBAAgB,KAAK,QAAQ,OAAO,IAAI,GAAG,KACjD,CAAC;AAGF,UADoB,MAAM,KAAK,mBAAmB,OAAO,IAAK,CAAC,EAAE,CAAC,EAChD;;;;;;;CAQpB,UACE,OACuB;AACvB,SAAO;GACL,OAAO,KAAK;GACZ,OAAO,EACL,OACD;GACD,YAAY,KAAK;GAClB;;;;;;;;;CAUH,MAAc,mBAAmB,MAA6B;AAC5D,SAAO,MACL,2FACA;GACE,QAAQ;GACR,SAAS;IACP,gBAAgB;IAChB,eAAe,UAAU,KAAK;IAC/B;GACD,MAAM,KAAK,UAAU,KAAK;GAC3B,CACF,CAAC,KAAK,OAAO,aAAa;GACzB,MAAM,gBACJ,MAAM,SAAS,MAAM;AAEvB,OAAI,UAAU,iBAAiB,cAAc,KAC3C,OAAM,IAAI,MAAM,GAAG,cAAc,KAAK,IAAI,cAAc,UAAU;AAGpE,UAAQ,cAAoC,OAAO,WAAW,KAC3D,EAAE,gBAAgB,UACpB;IACD"}