Why use CodeMind
A local Python RAG assistant for asking questions about local files and web pages.
Keep document Q&A local
CodeMind is useful for learning how a RAG pipeline works while keeping model inference and the vector index on your own machine.
The repository is intentionally a learning project: its chunking, retrieval, model adapter, and chat loop are visible code that you can inspect and tune rather than a black-box hosted service.
Use Ollama if you already have local models, or use the llama.cpp path when you want to work with GGUF models in the project. Changing model size trades answer speed and capability against memory and disk requirements.
It is not a hosted web application or a production security boundary. Do not ingest confidential files on a shared or untrusted machine, and remember that a local model can still produce incorrect answers.