Episode three tackles a single, well-scoped change: turning on sensitive data capture so full prompts and model responses get logged alongside the telemetry from the previous episode. It’s a small code change with a big tradeoff to think through. Having the actual message content in your logs is genuinely useful for debugging, auditing, and compliance, but it comes with real risks if prompts might contain personal or sensitive information, and it can meaningfully increase logging costs at high volumes. The video shows how this data lands in a new Gen AI content log in Application Insights, separate from the usual dependency log, and how to join the two by operation ID to get a complete picture of a request.
The code for this sample is here: https://github.com/michaelstephensonuk/azure-ai-samples/tree/main/03.helloworld.logging.with.message.body
