feature req: Show inference streams for debug purposes

[goktug7913]

Sometimes the models i use get into various failure modes, and with some clients i have no way of checking what's going on with the inference.

It would be extremely nice if there was a menu to see what the model is doing, what prompt hit etc. inside from oMLX (web or app UI)

I could probably do a PR for this but not sure if it'd get merged.