We read that local agents are becoming more and more useful across the board so we started experimenting with building models that enable. Ended up using this ourselves. Here is the git repo if you want to give it a go -> https://github.com/distil-labs/gitara
[edit] have you tried with a smaller model? I'm afraid a 3B model will brick my rusty laptop
sorry, missed that part, the accuracy drop-off isn't that bad!