Skip to main content

Bedrock

Amazon Bedrock is a fully managed service that makes base models from Amazon and third-party model providers accessible through an API.

When this documentation was written, Bedrock supports one model for text embeddings, the Titan Embeddings G1 - Text model (amazon.titan-embed-text-v1). This model supports text retrieval, semantic similarity, and clustering. The maximum input text is 8K tokens and the maximum output vector length is 1536.

Setup

To use this embedding, please ensure you have the Bedrock runtime client installed in your project.

npm i @aws-sdk/client-bedrock-runtime@^3.422.0
npm install @langchain/community
tip

We're unifying model params across all packages. We now suggest using model instead of modelName, and apiKey for API keys.

Usage

The BedrockEmbeddings class uses the AWS Bedrock API to generate embeddings for a given text. It strips new line characters from the text as recommended.

/* eslint-disable @typescript-eslint/no-non-null-assertion */
import { BedrockEmbeddings } from "@langchain/community/embeddings/bedrock";

const embeddings = new BedrockEmbeddings({
region: process.env.BEDROCK_AWS_REGION!,
credentials: {
accessKeyId: process.env.BEDROCK_AWS_ACCESS_KEY_ID!,
secretAccessKey: process.env.BEDROCK_AWS_SECRET_ACCESS_KEY!,
},
model: "amazon.titan-embed-text-v1", // Default value
});

const res = await embeddings.embedQuery(
"What would be a good company name a company that makes colorful socks?"
);
console.log({ res });

API Reference:

Configuring the Bedrock Runtime Client

You can pass in your own instance of the BedrockRuntimeClient if you want to customize options like credentials, region, retryPolicy, etc.

import { BedrockRuntimeClient } from "@aws-sdk/client-bedrock-runtime";
import { BedrockEmbeddings } from "langchain/embeddings/bedrock";

const client = new BedrockRuntimeClient({
region: "us-east-1",
credentials: getCredentials(),
});

const embeddings = new BedrockEmbeddings({
client,
});

Was this page helpful?


You can also leave detailed feedback on GitHub.