New
You can now attach evaluations to any log, trace, or session in Maxim. Instead of only evaluating new logs after setup, you can now run evals on historical logs – even if online evals weren’t previously configured. This enables you to analyze past data and gain granular, node-level insights into agent performance. Key highlights:
- Run evals on past logs by simply selecting those traces/sessions and adding evaluators based on the key metrics you wish to track.
- This helps you track agent performance over an extended timeframe to get a clear, metric-driven view of quality improvements or degradations.
- Filter logs by failure scenarios and re-run or attach additional evals for iterative debugging and deeper analysis.