ROZUM — a private AI assistant for your documents (RAG)

Engineers drown in documentation: tech cards, reports, instructions, diagrams. ROZUM is my desktop assistant that reads your files and answers from them, with links to the source.

How RAG works in plain words

1. You upload documents by section (PDF, DOCX, Excel).
2. The system splits them into chunks, computes embeddings and stores them in a vector database.
3. On a query it finds the most relevant chunks and feeds them to the model as context.
4. The answer comes with citations — you can see where it came from, down to the page.

Why local

The most important thing is privacy. Documents never leave the computer. There are two modes:

• Local GPU (Ollama) — fully offline, zero leakage.
• Cloud (Groq) — when you need it faster/smarter on harder queries.

Stack

Python + FastAPI, a pywebview interface (a native window), SQLite for metadata, local embeddings. Search across hundreds of pages — in seconds instead of manual scrolling.

Who needs this

Anyone who works with large sets of documents and does not want to hand them to someone else's cloud: engineers, lawyers, accountants, researchers.