I usually bounce between LM Studio for casual chat, AnythingLLM for RAG workflows, llama.cpp for cutting-edge features, and a ...