replica-omnisciente/dirac/docs/providers

Provider-Specific Settings

This document details the environment variables and configuration options for specific AI providers in Dirac.

AWS Bedrock

Use Bedrock by setting AWS credentials and region. When AWS_ACCESS_KEY_ID or AWS_BEDROCK_MODEL is present, Dirac automatically switches to the Bedrock provider.

Environment Variables

• AWS_ACCESS_KEY_ID — AWS access key
• AWS_SECRET_ACCESS_KEY — AWS secret key
• AWS_SESSION_TOKEN — session token (for temporary credentials)
• AWS_REGION — AWS region (e.g. us-east-1). Note: AWS_REGION alone will not trigger an automatic switch to Bedrock.
• AWS_BEDROCK_MODEL — model ID for both act and plan modes (e.g. us.anthropic.claude-sonnet-4-6)
• AWS_BEDROCK_MODEL_ACT — model ID for act mode only
• AWS_BEDROCK_MODEL_PLAN — model ID for plan mode only

Usage Example

Works seamlessly with aws-vault:

AWS_REGION=us-east-1 AWS_BEDROCK_MODEL=us.anthropic.claude-sonnet-4-6 \
  aws-vault exec my-profile -- dirac "your task"

Note: Newer Claude models on Bedrock (Sonnet 4.6+) require a cross-region inference profile prefix (us., eu., ap.). See the AWS docs for supported model IDs.

Google Cloud Vertex AI

Use Vertex AI by setting the following environment variables. When GOOGLE_CLOUD_PROJECT or GCP_PROJECT is present, Dirac automatically switches to the Vertex provider.

Environment Variables

• GOOGLE_CLOUD_PROJECT or GCP_PROJECT — Your Google Cloud project ID.
• GOOGLE_CLOUD_LOCATION or GOOGLE_CLOUD_REGION — The region for Vertex AI (e.g. us-central1).

Authentication

Dirac uses the default Google Cloud authentication chain. Ensure you have authenticated using the Google Cloud CLI:

gcloud auth application-default login