An AI practitioner is using a Bedrock base model to summarize customer service chat sessions and wants to keep logs of model inputs and outputs for monitoring. Which approach should they use to capture those invocation logs?

Enabling invocation logging in Amazon Bedrock is the correct approach because Bedrock provides native integration with Amazon CloudWatch Logs and Amazon S3 for logging model inputs, outputs, and other invocation details. This feature is specifically designed for monitoring and auditing Bedrock model usage. Configuring AWS CloudTrail is incorrect because CloudTrail primarily logs API calls and account activity, not the detailed input/output payloads of model invocations. AWS Audit Manager is incorrect as it's used for continuous auditing and compliance, not for capturing operational logs of model interactions. Configuring model invocation logging in Amazon EventBridge is incorrect because EventBridge is a serverless event bus that routes events between AWS services and your applications, not a logging destination for model invocation details.