Skip to main content

Cohere Rerank

对文档进行重新排序可以极大地改进任何 RAG 应用和文档检索系统。

¥Reranking documents can greatly improve any RAG application and document retrieval system.

从高层次来看，rerank API 是一种语言模型，它分析文档并根据其与给定查询的相关性对其进行重新排序。

¥At a high level, a rerank API is a language model which analyzes documents and reorders them based on their relevance to a given query.

Cohere 提供了用于对文档进行重新排序的 API。在本例中，我们将向你展示如何使用它。

¥Cohere offers an API for reranking documents. In this example we'll show you how to use it.

设置​

¥Setup

tip

请参阅 此部分包含有关安装集成包的一般说明。

¥See this section for general instructions on installing integration packages.

npm install @langchain/cohere @langchain/core
import { CohereRerank } from "@langchain/cohere";
import { Document } from "@langchain/core/documents";

const query = "What is the capital of the United States?";
const docs = [
  new Document({
    pageContent:
      "Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.",
  }),
  new Document({
    pageContent:
      "The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.",
  }),
  new Document({
    pageContent:
      "Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.",
  }),
  new Document({
    pageContent:
      "Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.",
  }),
  new Document({
    pageContent:
      "Capital punishment (the death penalty) has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.",
  }),
];

const cohereRerank = new CohereRerank({
  apiKey: process.env.COHERE_API_KEY, // Default
  model: "rerank-english-v2.0",
});

const rerankedDocuments = await cohereRerank.rerank(docs, query, {
  topN: 5,
});

console.log(rerankedDocuments);
/**
[
  { index: 3, relevanceScore: 0.9871293 },
  { index: 1, relevanceScore: 0.29961726 },
  { index: 4, relevanceScore: 0.27542195 },
  { index: 0, relevanceScore: 0.08977329 },
  { index: 2, relevanceScore: 0.041462272 }
]
 */

API Reference:

这里，我们可以看到 .rerank() 方法仅返回文档的索引（与输入文档的索引匹配）及其相关性得分。

¥Here, we can see the .rerank() method returns just the index of the documents (matching the indexes of the input documents) and their relevancy scores.

如果我们想从方法本身返回文档，我们可以使用 .compressDocuments() 方法。

¥If we'd like to have the documents returned from the method itself, we can use the .compressDocuments() method.

import { CohereRerank } from "@langchain/cohere";
import { Document } from "@langchain/core/documents";

const query = "What is the capital of the United States?";
const docs = [
  new Document({
    pageContent:
      "Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.",
  }),
  new Document({
    pageContent:
      "The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.",
  }),
  new Document({
    pageContent:
      "Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.",
  }),
  new Document({
    pageContent:
      "Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.",
  }),
  new Document({
    pageContent:
      "Capital punishment (the death penalty) has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.",
  }),
];

const cohereRerank = new CohereRerank({
  apiKey: process.env.COHERE_API_KEY, // Default
  topN: 3, // Default
  model: "rerank-english-v2.0",
});

const rerankedDocuments = await cohereRerank.compressDocuments(docs, query);

console.log(rerankedDocuments);
/**
[
  Document {
    pageContent: 'Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.',
    metadata: { relevanceScore: 0.9871293 }
  },
  Document {
    pageContent: 'The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.',
    metadata: { relevanceScore: 0.29961726 }
  },
  Document {
    pageContent: 'Capital punishment (the death penalty) has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.',
    metadata: { relevanceScore: 0.27542195 }
  }
]
 */

API Reference:

从结果中，我们可以看到它返回了排名前 3 位的文档，并为每个文档分配了一个 relevanceScore。

¥From the results, we can see it returned the top 3 documents, and assigned a relevanceScore to each.

不出所料，relevanceScore 得分最高的文档是引用华盛顿特区的文档，得分为 98.7%！

¥As expected, the document with the highest relevanceScore is the one that references Washington, D.C., with a score of 98.7%!

与 CohereClient 结合使用​

¥Usage with CohereClient

如果你在 Azure、AWS Bedrock 或独立实例上使用 Cohere，则可以使用 CohereClient 和你的端点创建 CohereRerank 实例。

¥If you are using Cohere on Azure, AWS Bedrock or a standalone instance you can use the CohereClient to create a CohereRerank instance with your endpoint.

import { CohereRerank } from "@langchain/cohere";
import { CohereClient } from "cohere-ai";
import { Document } from "@langchain/core/documents";

const query = "What is the capital of the United States?";
const docs = [
  new Document({
    pageContent:
      "Carson City is the capital city of the American state of Nevada. At the 2010 United States Census, Carson City had a population of 55,274.",
  }),
  new Document({
    pageContent:
      "The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean that are a political division controlled by the United States. Its capital is Saipan.",
  }),
  new Document({
    pageContent:
      "Charlotte Amalie is the capital and largest city of the United States Virgin Islands. It has about 20,000 people. The city is on the island of Saint Thomas.",
  }),
  new Document({
    pageContent:
      "Washington, D.C. (also known as simply Washington or D.C., and officially as the District of Columbia) is the capital of the United States. It is a federal district. The President of the USA and many major national government offices are in the territory. This makes it the political center of the United States of America.",
  }),
  new Document({
    pageContent:
      "Capital punishment (the death penalty) has existed in the United States since before the United States was a country. As of 2017, capital punishment is legal in 30 of the 50 states. The federal government (including the United States military) also uses capital punishment.",
  }),
];

const client = new CohereClient({
  token: process.env.COHERE_API_KEY,
  environment: "<your-cohere-deployment-url>", // optional
  // other params
});

const cohereRerank = new CohereRerank({
  client, // apiKey will be ignored even if provided
  model: "rerank-english-v2.0",
});

const rerankedDocuments = await cohereRerank.rerank(docs, query, {
  topN: 5,
});

console.log(rerankedDocuments);

/*
  [
    { index: 3, relevanceScore: 0.9871293 },
    { index: 1, relevanceScore: 0.29961726 },
    { index: 4, relevanceScore: 0.27542195 },
    { index: 0, relevanceScore: 0.08977329 },
    { index: 2, relevanceScore: 0.041462272 }
  ]
*/

API Reference: