Question

A team runs a customer-support chatbot built with API Gateway, Lambda, and Amazon Bedrock. Occasionally a user reports a garbled or incomplete answer, but engineers cannot reconstruct which model invocation, retrieved documents, and prompt version produced that specific response. They want an observability approach that lets them trace a single end-user request across all components and inspect the exact Bedrock input/output for that request. Which approach best meets this requirement with the least custom instrumentation?

Accepted Answer

Enable AWS X-Ray active tracing on API Gateway and Lambda, propagate the X-Ray trace ID as a correlation identifier, and enable Bedrock model invocation logging to CloudWatch/S3 so the request/response payloads can be correlated by that identifier.

Answer

Turn on CloudWatch Container Insights and rely on aggregate p99 latency dashboards to identify the affected time window, then manually re-run the user's query.

Answer

Increase the Lambda log retention to 90 days and grep CloudWatch Logs Insights for the user's message text to find the matching invocation.

Answer

Enable Bedrock CloudWatch metrics (Invocations, InputTokenCount) and create an anomaly-detection alarm to catch degraded responses automatically.