§06·2025Live demo
RAG-QA App
Upload a PDF, ask questions, get cited answers.
Summary
A production RAG service: PDF upload, FAISS vector store, source-attributed answers via gpt-3.5-turbo. Frontend on Streamlit Cloud, FastAPI backend on Render, CI on GitHub Actions.
The problem
Most RAG tutorials stop at the notebook. This one ships to production with two services, real CI, and a public URL.
Approach
- 01Streamlit handles upload + chat UI.
- 02FastAPI service does chunking + embedding + retrieval.
- 03FAISS holds vectors per session; OpenAI runs the LLM.
Highlights
- ·Up to 200-page PDFs
- ·Source attribution in answers
- ·Decoupled frontend + backend, both deployed
- ·Pytest + GitHub Actions CI