sowmiyan-s/GUARD-RAG ? reverse-engineered prompt

Build me a private, offline document chat app called GuardRAG.

I want a Python package that lets me upload local files like PDFs, Word docs, text, markdown, CSVs, JSON, logs, and Python files, then ask questions about them in a chat interface. It should run fully on my machine with local models through Ollama, use a local vector store for search, and always answer with source citations and page numbers when possible.

Please include strong safety checks, like redacting sensitive info, blocking obvious secrets, and protecting against prompt injection in uploaded files. I also want live streaming responses, an audit log, a simple API for programmatic use, and a command line mode for batch use. If it makes sense, add a Docker setup and a clean README with quick start instructions. Use the current docs online if you need to confirm anything.