Local Code Chatbot Running on 2GB RAM

There is even some untapped headroom, as they quantized to Q4 instead of using K-quant.

Not to speak of the potential hooked up to a vector db, swapping out LORAs for different languages and such.