AWS Bedrock Integration#

The NeMo Agent toolkit supports integration with multiple LLM providers, including AWS Bedrock. This documentation provides a comprehensive guide on how to integrate AWS Bedrock models into your NeMo Agent toolkit workflow. To view the full list of supported LLM providers, run nat info components -t llm_provider.

Configuration#

Prerequisites#

Before integrating AWS Bedrock, ensure you have:

  • Set up AWS credentials by configuring AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY

  • For detailed setup instructions, refer to the AWS Bedrock setup guide

Example Configuration#

Add the AWS Bedrock LLM configuration to your workflow config file. Make sure the region_name matches the region of your AWS account, and the credentials_profile_name matches the field in your credential file:

llms:
  aws_bedrock_llm:
    _type: aws_bedrock
    model_name: meta.llama3-3-70b-instruct-v1:0
    temperature: 0.0
    max_tokens: 1024
    region_name: us-east-2
    credentials_profile_name: default

Configurable Options#

  • model_name: The name of the AWS Bedrock model to use (required)

  • temperature: Controls randomness in the output (0.0 to 1.0, default: 0.0)

  • max_tokens: Maximum number of tokens to generate (must be > 0, default: 1024)

  • context_size: Maximum number of tokens for context (must be > 0, default: 1024, required for LlamaIndex)

  • region_name: AWS region where your Bedrock service is hosted (default: “None”)

  • base_url: Custom Bedrock endpoint URL (default: None, needed if you don’t want to use the default us-east-1 endpoint)

  • credentials_profile_name: AWS credentials profile name from ~/.aws/credentials or ~/.aws/config files (default: None)

Usage in Workflow#

Reference the AWS Bedrock LLM in your workflow configuration:

workflow:
  _type: react_agent
  llm_name: aws_bedrock_llm
  # ... other workflow configurations