跳到主要内容
兼容性仅在 Node.js 上可用。
OpenSearchElasticsearch 的一个分支,与 Elasticsearch API 完全兼容。在此处阅读更多关于其对近似最近邻居的支持 此处 LangChain.js 接受 @opensearch-project/opensearch 作为 OpenSearch 向量存储的客户端。

设置

有关安装 LangChain 软件包的一般说明,请参阅此部分
npm
npm install -S @langchain/openai @langchain/core @opensearch-project/opensearch
您还需要运行一个 OpenSearch 实例。您可以使用官方 Docker 镜像开始使用。您还可以在此处找到一个示例 docker-compose 文件。

索引文档

import { Client } from "@opensearch-project/opensearch";
import { Document } from "@langchain/classic/document";
import { OpenAIEmbeddings } from "@langchain/openai";
import { OpenSearchVectorStore } from "@langchain/community/vectorstores/opensearch";

const client = new Client({
  nodes: [process.env.OPENSEARCH_URL ?? "http://127.0.0.1:9200"],
});

const docs = [
  new Document({
    metadata: { foo: "bar" },
    pageContent: "opensearch is also a vector db",
  }),
  new Document({
    metadata: { foo: "bar" },
    pageContent: "the quick brown fox jumped over the lazy dog",
  }),
  new Document({
    metadata: { baz: "qux" },
    pageContent: "lorem ipsum dolor sit amet",
  }),
  new Document({
    metadata: { baz: "qux" },
    pageContent:
      "OpenSearch is a scalable, flexible, and extensible open-source software suite for search, analytics, and observability applications",
  }),
];

await OpenSearchVectorStore.fromDocuments(docs, new OpenAIEmbeddings(), {
  client,
  indexName: process.env.OPENSEARCH_INDEX, // Will default to `documents`
});

查询文档

import { Client } from "@opensearch-project/opensearch";
import { VectorDBQAChain } from "@langchain/classic/chains";
import { OpenAIEmbeddings } from "@langchain/openai";
import { OpenAI } from "@langchain/openai";
import { OpenSearchVectorStore } from "@langchain/community/vectorstores/opensearch";

const client = new Client({
  nodes: [process.env.OPENSEARCH_URL ?? "http://127.0.0.1:9200"],
});

const vectorStore = new OpenSearchVectorStore(new OpenAIEmbeddings(), {
  client,
});

/* Search the vector DB independently with meta filters */
const results = await vectorStore.similaritySearch("hello world", 1);
console.log(JSON.stringify(results, null, 2));
/* [
    {
      "pageContent": "Hello world",
      "metadata": {
        "id": 2
      }
    }
  ] */

/* Use as part of a chain (currently no metadata filters) */
const model = new OpenAI();
const chain = VectorDBQAChain.fromLLM(model, vectorStore, {
  k: 1,
  returnSourceDocuments: true,
});
const response = await chain.call({ query: "What is opensearch?" });

console.log(JSON.stringify(response, null, 2));
/*
  {
    "text": " Opensearch is a collection of technologies that allow search engines to publish search results in a standard format, making it easier for users to search across multiple sites.",
    "sourceDocuments": [
      {
        "pageContent": "What's this?",
        "metadata": {
          "id": 3
        }
      }
    ]
  }
  */

以编程方式连接这些文档到 Claude、VSCode 等,通过 MCP 获取实时答案。
© . This site is unofficial and not affiliated with LangChain, Inc.