MLflow Tracing
MLflow’s tracing interface for recording and inspecting steps inside generative AI applications.
Published by MLflow Project
How to use it
Record model calls, retrieval, and application logic as spans in one trace. Give each span a consistent name so you can compare latency and errors before and after a code or prompt change.
Start MLflow locally, run a small workload, and inspect the trace in its UI. Attach a trace ID to evaluation results to make failures reproducible.
- Compatibility
- Python, MLflow Tracking
- Monitoring areas
- LLM tracing, Experiment tracking, Trace-linked evaluation
- Access
- Local Python library and tracking UI. Remote storage is optional.