amazon-bedrock-agentcore-samples/02-use-cases/04-SRE-agent/docs/configuration.md

Configuration

Environment Variables

The SRE Agent uses environment variables for sensitive configuration values. Create a .env file in the sre_agent/ directory with the following required variables:

# Required: API key for Claude model access
# For Anthropic direct access:
ANTHROPIC_API_KEY=sk-ant-api-key-here

# For Amazon Bedrock access:
AWS_DEFAULT_REGION=us-east-1
AWS_PROFILE=your-profile-name  # Or use AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY

# Required: AgentCore Gateway authentication
GATEWAY_ACCESS_TOKEN=your-gateway-token-here  # Generated by gateway setup

# Optional: Debugging and logging
LOG_LEVEL=INFO  # Options: DEBUG, INFO, WARNING, ERROR
DEBUG=false     # Enable debug mode for verbose output

Note: The SRE Agent looks for the .env file in the sre_agent/ directory, not the project root. This allows for modular configuration management.

Agent Configuration

The agent behavior is configured through sre_agent/config/agent_config.yaml. This file defines the mapping between agents and their available tools, as well as LLM parameters:

# Agent to tool mapping
agents:
  kubernetes_agent:
    name: "Kubernetes Infrastructure Agent"
    description: "Specializes in Kubernetes operations and troubleshooting"
    tools:
      - get_pod_status
      - get_deployment_status
      - get_cluster_events
      - get_resource_usage
      - get_node_status

  logs_agent:
    name: "Application Logs Agent"
    description: "Expert in log analysis and pattern detection"
    tools:
      - search_logs
      - get_error_logs
      - analyze_log_patterns
      - get_recent_logs
      - count_log_events

  metrics_agent:
    name: "Performance Metrics Agent"
    description: "Analyzes performance metrics and trends"
    tools:
      - get_performance_metrics
      - get_error_rates
      - get_resource_metrics
      - get_availability_metrics
      - analyze_trends

  runbooks_agent:
    name: "Operational Runbooks Agent"
    description: "Provides operational procedures and guides"
    tools:
      - search_runbooks
      - get_incident_playbook
      - get_troubleshooting_guide
      - get_escalation_procedures
      - get_common_resolutions

# Global tools available to all agents
global_tools:
  - x-amz-bedrock-agentcore-search  # AgentCore search tool
  
# Gateway configuration
gateway:
  uri: "https://your-gateway-url.com"  # Updated during setup

Gateway Configuration

The AgentCore Gateway is configured through gateway/config.yaml. This configuration is managed by the setup scripts but can be customized:

# AgentCore Gateway Configuration Template
# Copy this file to config.yaml and update with your environment-specific settings

# AWS Configuration
account_id: "YOUR_ACCOUNT_ID"
region: "us-east-1"
role_name: "YOUR_ROLE_NAME"
endpoint_url: "https://bedrock-agentcore-control.us-east-1.amazonaws.com"
credential_provider_endpoint_url: "https://us-east-1.prod.agent-credential-provider.cognito.aws.dev"

# Cognito Configuration
user_pool_id: "YOUR_USER_POOL_ID"
client_id: "YOUR_CLIENT_ID"

# S3 Configuration
s3_bucket: "your-agentcore-schemas-bucket"
s3_path_prefix: "devops-multiagent-demo"  # Path prefix for OpenAPI schema files

# Provider Configuration
# This ARN is automatically generated by create_gateway.sh when it runs create_credentials_provider.py
provider_arn: "arn:aws:bedrock-agentcore:REGION:ACCOUNT_ID:token-vault/default/apikeycredentialprovider/YOUR_PROVIDER_NAME"

# Gateway Configuration
gateway_name: "MyAgentCoreGateway"
gateway_description: "AgentCore Gateway for API Integration"

# Target Configuration
target_description: "S3 target for OpenAPI schema"