I tried a couple of times with Jen ai and local llama, but somehow does not work that well for me.

But at the same time i have a 9070xt, so, not exactly optimal