So I wasted a year doing this with PosterChan, the best localAI solution.
You won't have any room on your GPU for context (your code) because the model will take up mode than your GPU because ots fucking huge. Your project is too large for it to read the code properly. Using a smaller model is even more retarded than the model you are using.
You are wasting your time. If you want to code properly like Claude at home, you need 128GB of vram, more better
