About This Project
Overview
NoteMind applies Retrieval-Augmented Generation to personal notes: a FastAPI + LangChain backend OCRs and embeds your notes into FAISS, then answers questions grounded in that indexed content via Gemini — turning a pile of notes into something you can actually query.
Motivation
My notes were scattered and unsearchable in any meaningful way, so I built a RAG pipeline that lets me ask questions against my own notes instead of scrolling through them.
Lessons Learned
Chunking strategy mattered far more than model choice for answer quality — bad chunk boundaries produced confidently wrong answers no matter how good the underlying LLM was.