Building a free, offline LLM “tutor” grounded in one university textbook — RAG, LoRA, or both? Sanity check wanted

In three linesDeveloper seeks to build a free offline AI tutor grounded in a university textbook. Proposed architecture: RAG as core component (chunking, embedding, retrieval with page/section citations) + optional LoRA for pedagogical style. Questions on model selection (Qwen, Gemma), handling complex structures (figures, equations), and packaging for non-technical users.

Summary generated by Claude — human-verified