Feature Request: Web UI for Viewing Evaluation Results

[Leo-Lok]

Hi team,

First of all, thank you for this excellent framework! I've been using EVA to evaluate voice agents for AI scenarios.

Currently, evaluation results are output as JSON files in debug_output/. While this works, I find it challenging to share and discuss results with colleagues.

Feature Request:

Would it be possible to add a web-based visualization interface (similar to what promptfoo provides with promptfoo view)?

Thanks!

[gabegma]

Hi!
Thank you for your interest in using eva!
Have you already seen our Streamlit analysis app? We use it to compare metrics across runs and samples and to investigate specific conversations. For specific conversations (Record detail), there are 2 tabs we use a lot; the first one to see the different metrics per turn, and the details of every metric at the top. The second one is the turn-taking analysis (last tab), where you can play the audio for specific turns and inspect latency.

Let us know if that's helpful!

[Leo-Lok]

Hi!
Thank you for your reply regarding the web UI.

I realized that I had checked out the "latest" branch instead of the main branch, which is why I missed the visualization feature. Thank you for pointing it out.

Best regards,
Leo