Advanced Error Handling and Recovery for Amazon Bedrock Agents (Python)

Advanced Error Handling and Recovery for Amazon Bedrock Agents (Python)

This guide covers advanced error handling and recovery patterns for building resilient and reliable Bedrock Agents.

1. Common Errors in Bedrock Agents

• ValidationException: Invalid input parameters.
• AccessDeniedException: Insufficient IAM permissions.
• ResourceNotFoundException: The specified agent, knowledge base, or action group does not exist.
• ThrottlingException: The request rate exceeds the limit.
• ServiceUnavailableException: An internal server error occurred.
• DependencyFailedException: An error occurred in a downstream service, such as a Lambda function or a knowledge base.

2. Error Handling Patterns

a. Retry with Exponential Backoff

Implement a retry mechanism with exponential backoff to handle transient errors like ThrottlingException and ServiceUnavailableException.

import time
from botocore.exceptions import ClientError

def invoke_agent_with_retry(agent_id, input_text, max_retries=3):
    for i in range(max_retries):
        try:
            response = bedrock_runtime.invoke_agent(
                agentId=agent_id,
                inputText=input_text,
            )
            return response
        except ClientError as e:
            if e.response['Error']['Code'] in ['ThrottlingException', 'ServiceUnavailableException']:
                time.sleep(2 ** i)
            else:
                raise

b. Dead-Letter Queues (DLQs)

Use a Dead-Letter Queue (DLQ) to capture and analyze failed agent invocations. You can configure a DLQ on your Lambda functions or use an Amazon SQS queue.

c. Custom Error Handling in Action Groups

Implement custom error handling in your action group Lambda functions to provide more meaningful error messages to the agent.

def lambda_handler(event, context):
    try:
        # Your action group logic here
        pass
    except Exception as e:
        return {
            'error': {
                'type': 'CustomError',
                'message': str(e)
            }
        }

3. Recovery Patterns

a. Checkpointing and Recovery

For long-running tasks, use the built-in checkpointing and recovery feature of AgentCore to save the agent’s state and resume the task from the last checkpoint in case of a failure.

b. Fallback Agents

Implement a fallback mechanism to a simpler agent or a human agent if the primary agent fails to complete a task.

try:
    response = primary_agent.invoke(request)
except Exception as e:
    response = fallback_agent.invoke(request)