The “Logs” page is essential, storing all captured interactions with detailed reports, especially focusing on Evaluations and their outcomes. In case of a failure, it highlights issues, failed checks, and offers in-depth explanations about what went wrong.

The explanations can be comprehended through two distinct modes:

  1. Classic mode - where you read the interaction and the issues separately.

  2. Visual mode - uses colored labels to point out issues in interactions, making it easier to understand. For example, it marks PIIs in purple.

You can also access raw data and explore a feature that suggests better prompts.