Is there any way to see if the Vercel AI SDK is using the cache?

[dschlabach]

I've set up a local cache similar to what you've done in packages/examples, and I think it's working, but given that it'll cost me if I'm wrong, is there a way to indicate in the eval tables if an answer is cached?

[mattpocock]

Neat idea, I'll add some metadata

[mattpocock]

A cheap way to do this (which is now done) is to show the durations more clearly in the traces view. That way you can see that if it took 10ms, it was likely locally cached.

[dschlabach]

Awesome, I'll add traces to my setup!

[mattpocock]

Also added a spot for tracking specific numbers of prompt and completion tokens: