Skip to main content

AWS Bedrock Setup

LlamaCloud supports AWS Bedrock as part of its multimodal AI capabilities. To enable AWS Bedrock support, update the configuration to turn on the Bedrock integration. This page guides you through the necessary configuration changes and permission settings to integrate AWS Bedrock into your deployment.

  • A valid AWS account
  • Access & Quota to the supported models:
    • For LlamaParse advanced parsing features:
      • Anthropic Claude Sonnet 3.7, specifically anthropic.claude-3-7-sonnet-20250219-v1:0. If this version is not available an alternative versionName can be configure through the environment variable BEDROCK_ANTHROPIC_SONNET_3_7_VERSION_NAME
        • Note: This model is required if you want to use the parse_page_with_lvm, parse_page_with_agent or parse_page_document_with_agent parsing modes.
      • Anthropic Claude Sonnet 3.5, specifically anthropic.claude-3-5-sonnet-20240620-v1:0. If this version is not available an alternative versionName can be configure through the environment variable BEDROCK_ANTHROPIC_SONNET_3_5_VERSION_NAME
        • Note: This model is required if you want to use the parse_page_with_lvm or parse_page_with_agent parsing modes.
    • For LlamaCloud Playground:
      • Anthropic Claude Sonnet 3.5 V2
      • Cohere Rerank 3.5 (Optional)
        • Note: Please check if this model is available in your region.

Connecting to Bedrock​

Below are two ways to configure a connection to AWS Bedrock:

We recommend creating a new IAM Role and Policy for LlamaCloud. You can then attach the role ARN as a service account annotation.

Below is a loose-goosey example policy to serve as a reference. Please tighten the policy as needed.

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "bedrock:InvokeModel",
        "bedrock:InvokeModelWithResponseStream",
        "bedrock:CreateModelInvocationJob"
      ],
      "Resource": [
        "arn:aws:bedrock:*:*:foundation-model/*",
        "arn:aws:bedrock:*:*:inference-profile/*"
      ]
    }
  ]
}

After creating a policy similar to the above, update the LlamaParse service accounts with the EKS annotation:

# Example for the backend service account. Repeat for each of the services listed above.
llamaParse:
  config:
    awsBedrock:
      enabled: true
  serviceAccount:
    annotations:
      eks.amazonaws.com/role-arn: arn:aws:iam::<account-id>:role/<role-name>

For more information, refer to the official AWS documentation about this topic.

IAM Role w/ Static AWS Credentials​

Create a user with a policy attached for Bedrock Model Invocation. Afterwards, configure the platform to use the AWS credentials of that user by setting the following values in your values.yaml file:

llamaparse:
  config:
    awsBedrock:
      enabled: true
      region: "<your-aws-region>"
      accessKeyId: "<your-access-key-id>"
      secretAccessKey: "<your-secret-access-key>"
      # Optional: If the default version is not available, you can specify an alternative versionName
      # LlamaParse will use the alternative versionName to invoke the model
      # We support 3.5 and 3.7 in different LlamaParse parsing modes.
      sonnet3_5ModelVersionName: "anthropic.claude-3-5-sonnet-20240620-v1:0"
      sonnet3_7ModelVersionName: "anthropic.claude-3-7-sonnet-20250219-v1:0"