NanoAi. Offline LLM with RAG for Android — run GGUF models locally on-device using llama.cpp