Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Tip
Optionally, install mlflow/skills for a deeper integration with MLflow.
Once your agent is instrumented, its traces are where you see what happened and find where things went wrong. Start in the MLflow UI: open your experiment and click the Traces tab.
Here is a quick tour of what you can do from there:
Ask in plain English. Click the sparkle (Filter with AI) in the Search box and describe the traces you want — for example, "Find traces that call the get_weather tool." MLflow turns your description into a filter over the trace list.
Open a trace to inspect its details end to end: the span tree, and each step's inputs, outputs, timing, and errors.
Analyze the trace. Click Analyze trace in the trace panel to have MLflow analyze the trace for issues.
Add feedback. Click Assess and record a rating or note on the trace (or an expectation — the correct answer).

Next steps
View traces in the Databricks MLflow UI — Everything the trace UI can do: search, filter, columns, and the trace detail view.
Detect and monitor issues across MLflow traces — Detect recurring issues across traces, fix them, and monitor the results.
Collect feedback and build datasets — Collect feedback and build evaluation datasets.