Skip to main content

Momento Vector Index (MVI)

MVI：您数据的最高产、最易用的无服务器向量索引。要开始使用 MVI，只需注册一个账户即可。无需处理基础设施、管理服务器，也无需担心扩展问题。MVI 是一项可自动扩展以满足您需求的服务。无论是在 Node.js、浏览器还是边缘环境中，Momento 都能为您提供支持。

要注册并访问 MVI，请访问 Momento 控制台。

配置​

  1. 在 Momento 控制台中注册一个 API 密钥。

  2. 为您的环境安装 SDK。

    2.1 对于 Node.js：

    npm install @gomomento/sdk

    2.2 对于 浏览器或边缘环境：

    npm install @gomomento/sdk-web

  3. 在运行代码前为 Momento 设置环境变量

    3.1 OpenAI

    export OPENAI_API_KEY=YOUR_OPENAI_API_KEY_HERE

    3.2 Momento

    export MOMENTO_API_KEY=YOUR_MOMENTO_API_KEY_HERE # https://console.gomomento.com

使用方法​

:::提示 请参阅安装集成包的一般说明部分。 :::

npm install @langchain/openai @langchain/community @langchain/core

使用 fromTexts 方法索引文档并进行搜索​

此示例演示了如何使用 fromTexts 方法实例化向量存储并索引文档。 如果索引不存在，则会创建它。如果索引已存在，则会将文档添加到现有索引中。

ids 是可选的；如果您省略它们，Momento 会为您生成 UUID。

import { MomentoVectorIndex } from "@langchain/community/vectorstores/momento_vector_index";
// For browser/edge, adjust this to import from "@gomomento/sdk-web";
import {
  PreviewVectorIndexClient,
  VectorIndexConfigurations,
  CredentialProvider,
} from "@gomomento/sdk";
import { OpenAIEmbeddings } from "@langchain/openai";
import { sleep } from "langchain/util/time";

const vectorStore = await MomentoVectorIndex.fromTexts(
  ["hello world", "goodbye world", "salutations world", "farewell world"],
  {},
  new OpenAIEmbeddings(),
  {
    client: new PreviewVectorIndexClient({
      configuration: VectorIndexConfigurations.Laptop.latest(),
      credentialProvider: CredentialProvider.fromEnvironmentVariable({
        environmentVariableName: "MOMENTO_API_KEY",
      }),
    }),
    indexName: "langchain-example-index",
  },
  { ids: ["1", "2", "3", "4"] }
);

// because indexing is async, wait for it to finish to search directly after
await sleep();

const response = await vectorStore.similaritySearch("hello", 2);

console.log(response);

/*
[
  Document { pageContent: 'hello world', metadata: {} },
  Document { pageContent: 'salutations world', metadata: {} }
]
*/

API Reference:

使用 fromDocuments 方法索引文档并进行搜索​

与上述类似，此示例演示了如何使用 fromDocuments 方法实例化向量存储并索引文档。 如果索引不存在，则会创建它。如果索引已存在，则会将文档添加到现有索引中。

使用 fromDocuments 可以将各种文档加载器与索引无缝连接起来。

import { MomentoVectorIndex } from "@langchain/community/vectorstores/momento_vector_index";
// For browser/edge, adjust this to import from "@gomomento/sdk-web";
import {
  PreviewVectorIndexClient,
  VectorIndexConfigurations,
  CredentialProvider,
} from "@gomomento/sdk";
import { OpenAIEmbeddings } from "@langchain/openai";
import { TextLoader } from "langchain/document_loaders/fs/text";
import { sleep } from "langchain/util/time";

// Create docs with a loader
const loader = new TextLoader("src/document_loaders/example_data/example.txt");
const docs = await loader.load();

const vectorStore = await MomentoVectorIndex.fromDocuments(
  docs,
  new OpenAIEmbeddings(),
  {
    client: new PreviewVectorIndexClient({
      configuration: VectorIndexConfigurations.Laptop.latest(),
      credentialProvider: CredentialProvider.fromEnvironmentVariable({
        environmentVariableName: "MOMENTO_API_KEY",
      }),
    }),
    indexName: "langchain-example-index",
  }
);

// because indexing is async, wait for it to finish to search directly after
await sleep();

// Search for the most similar document
const response = await vectorStore.similaritySearch("hello", 1);

console.log(response);
/*
[
  Document {
    pageContent: 'Foo\nBar\nBaz\n\n',
    metadata: { source: 'src/document_loaders/example_data/example.txt' }
  }
]
*/

API Reference:

从现有集合中搜索​

import { MomentoVectorIndex } from "@langchain/community/vectorstores/momento_vector_index";
// For browser/edge, adjust this to import from "@gomomento/sdk-web";
import {
  PreviewVectorIndexClient,
  VectorIndexConfigurations,
  CredentialProvider,
} from "@gomomento/sdk";
import { OpenAIEmbeddings } from "@langchain/openai";

const vectorStore = new MomentoVectorIndex(new OpenAIEmbeddings(), {
  client: new PreviewVectorIndexClient({
    configuration: VectorIndexConfigurations.Laptop.latest(),
    credentialProvider: CredentialProvider.fromEnvironmentVariable({
      environmentVariableName: "MOMENTO_API_KEY",
    }),
  }),
  indexName: "langchain-example-index",
});

const response = await vectorStore.similaritySearch("hello", 1);

console.log(response);
/*
[
  Document {
    pageContent: 'Foo\nBar\nBaz\n\n',
    metadata: { source: 'src/document_loaders/example_data/example.txt' }
  }
]
*/

API Reference:

相关内容​

Was this page helpful?


You can also leave detailed feedback on GitHub.