Ability to edit agent #592

For versioning, having different metrics/ratings for each version would help users distinguish and choose. But then the issue is that data for each version will be sparse as compared to if we keep it agent wise ratings. Something to think about, how we can differentiate versions for users, while maintaining rich data and ratings.
Some options:
-can show version popularity -> 50% of users on v2.4.1 for example
-keep minimum thresholds to separate versions -> must have a certain amount of data
-give priority/value to recent or current versions/ratings/data)

All have drawbacks, auto-updating and new users will increase new version bias, hard thresholds might be poorly chosen. Should we target a hybrid approach?

Also, for updating, we can add a toggle for auto-update with update notifications so users can avoid manual updating.

Why not use our eval system to evaluate each version and display the scorecard for each so that users can know how/if it's improved? And, I think there should be no concept of auto update, we should let the user do it manually if they wish and maybe give them a reminder that there is a new version out.

The issue of sparse data remains, unless traces are produced artificially to compensate. But this will be far more negligible with a larger consumer base.
I see your point with the auto-update, maybe a mass update with multiple selection/deselection instead. Without it, managing and updating 20-30 popular and frequently updated agents can become a hassle.
Reminders are good, they can add a developer given description of the changes.