This is great news. Given the proliferation of solid local models, it would be cool if llamafile had a way to build your own custom versions with the model of your choice.