DocPy

RAG-powered Q&A for Python library documentation. Ask natural-language questions about any installed Python library and get answers with citations to specific functions and classes.

🚧 Status: In active development (Phase 0 complete)

Stack

• Gemini API — LLM + embeddings
• ChromaDB — local vector store
• Streamlit — web UI
• Python inspect — extracts symbols from installed libraries

Setup

curl -LsSf https://astral.sh/uv/install.sh | sh
git clone https://github.com/YOUR_USERNAME/docpy.git
cd docpy
uv sync
uv run streamlit run app.py

Get a free Gemini API key at https://aistudio.google.com/apikey, then copy .streamlit/secrets.toml.example to .streamlit/secrets.toml and add your key.

Deploy a public link (Streamlit Community Cloud)

This gives you a URL like https://docpy-yourname.streamlit.app you can open on any phone or share — no local server needed.

1. Push this repo to GitHub (already done).
2. Go to https://share.streamlit.io and sign in with GitHub.
3. Click Create app → Deploy a public app from GitHub.
4. Set:
   • Repository: Sebahstian/docpy
   • Branch: the branch you want to deploy
   • Main file path: app.py
5. Open Advanced settings → Secrets and paste your key in TOML form:

   GEMINI_API_KEY = "your-real-key"

6. Click Deploy. First build takes a few minutes (it installs requirements.txt).

Once it's live, index a library from the sidebar (requests is preloaded) and ask away. Note: you can only query libraries that are installed on the server — they must appear in requirements.txt.