AWS Bedrock Provider

AWS Bedrock provides access to foundation models from Anthropic, Meta, Mistral, Cohere, AI21, Amazon, and Stability AI through the Converse API.

Installation

npm install @aeye/aws

The AWS SDK v3 is included as a dependency.

Configuration

import { AWSBedrockProvider } from '@aeye/aws';

const aws = new AWSBedrockProvider({
  region: 'us-east-1',  // optional, defaults to AWS_REGION env var
  credentials: {         // optional, uses AWS credential chain by default
    accessKeyId: '...',
    secretAccessKey: '...',
    sessionToken: '...',  // for temporary credentials
  },
  modelPrefix: 'us.',   // optional, for cross-region inference
  defaultModels: {
    chat: 'anthropic.claude-3-5-sonnet-20241022-v2:0',
    imageGenerate: 'stability.stable-diffusion-xl-v1',
    embedding: 'amazon.titan-embed-text-v1',
  },
  hooks: {
    chat: {
      beforeRequest: (request, command, ctx) => {},
      afterRequest: (request, command, response, ctx) => {},
    },
  },
});

Credential Chain

AWS Bedrock automatically discovers credentials in this order:

1. Explicit credentials in config
2. Environment variables (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY)
3. Shared credentials file (~/.aws/credentials)
4. IAM roles (EC2, ECS, Lambda)
5. SSO (aws sso login)

Most setups need no explicit credentials:

const aws = new AWSBedrockProvider({
  region: 'us-east-1',
  // credentials discovered automatically
});

Supported Model Families

Family	Models	Capabilities
Anthropic	Claude 3.5 Sonnet, Claude 3 Haiku/Sonnet/Opus	Chat, Vision, Tools, Streaming, Reasoning
Meta	Llama 3, 3.1, 3.2	Chat, Streaming
Mistral	Mistral 7B, Mixtral, Mistral Large	Chat, Streaming
Cohere	Command R, Command R+	Chat, Streaming
AI21	Jurassic-2	Chat
Stability AI	SDXL, SD3	Image Generation
Amazon	Titan Text Embeddings	Embeddings

Cross-Region Inference

Access models in other regions using a model prefix:

const aws = new AWSBedrockProvider({
  region: 'us-east-1',
  modelPrefix: 'us.', // routes to US-based endpoints
});

Model Family Configuration

Enable/disable specific model families and customize model ID mappings:

const aws = new AWSBedrockProvider({
  region: 'us-east-1',
  modelFamilies: {
    anthropic: {
      enabled: true,
      modelIdMap: {
        'claude-3-sonnet': 'anthropic.claude-3-sonnet-20240229-v1:0',
      },
    },
    meta: { enabled: false }, // disable Llama models
  },
});

Usage

const ai = AI.with()
  .providers({ aws })
  .create();

// Chat with Claude on Bedrock
const response = await ai.chat.get(
  { messages },
  { metadata: { model: 'anthropic.claude-3-5-sonnet-20241022-v2:0' } }
);

// Image generation with Stability AI
const image = await ai.image.generate.get({
  prompt: 'A mountain landscape',
});

// Embeddings with Titan
const embeddings = await ai.embed.get({
  texts: ['Hello world'],
});

Error Types

import {
  AWSError,
  AWSAuthError,
  AWSRateLimitError,
  AWSQuotaError,
  AWSContextWindowError,
} from '@aeye/aws';

Model Access

As of 2025, Bedrock models auto-enable on first invocation in commercial regions. For Anthropic models, first-time users may need to submit use case details through the AWS console.