Getting good traces to see how your LLM app is performing is super important

Published: August 15th 2025

When building out AI Agent apps, you want to make as easy as possible to review how they are working behind the scenes

Create nice UI's that allow the software and business team to see what is going on

To see what Tools are called, what the Tool results are, and what responses the user receive.

Make it easy to see how the flow of a multi-turn conversation plays out.

And for the UI, don't just dump JSON for people to look at.

Make it nice and easy to look at.

→ Want to display some SQL? Show it in a code block → Have tabular data? Show it in a table → Displaying the chatbots response? Format it in markdown, just like the user would see it

Reviewing conversation traces, and more importantly, where they go wrong, is the foundation for building a well performing AI chatbot assistant.

Making those traces as easy and fun to review as possible, for both the development and business team, is going to make that process much more fluid,

Which means that you might actually do it!

.. and keep improving your bot over time