doc-qa-ai

RAG-powered document Q&A — upload any PDF and ask questions in plain language. Runs fully free with a Groq key and local embeddings.

How it works

1. Upload a PDF (up to 10 MB) or paste text
2. Text is chunked and embedded locally using sentence-transformers (all-MiniLM-L6-v2) — no API call, no cost
3. Embeddings are indexed in FAISS for fast cosine similarity search
4. Your question is embedded the same way and matched against the index
5. Top-4 chunks are passed to Groq (Llama 3.3 70B) for a grounded answer with source citations

Stack

Layer	Choice
LLM	Groq / Llama 3.3 70B (free tier)
Embeddings	sentence-transformers all-MiniLM-L6-v2 (local, no API key)
Vector store	FAISS (CPU)
UI	Streamlit
PDF parsing	pypdf
Alt provider	OpenAI (switchable via sidebar)

Setup

git clone https://github.com/biswajeetdev/doc-qa-ai
cd doc-qa-ai
pip install -r requirements.txt
cp .env.example .env        # add your GROQ_API_KEY
streamlit run app.py

Get a free Groq key (no credit card) at console.groq.com/keys.

The sentence-transformers model (~22 MB) downloads automatically on first run.

Security notes

• Chunk text is HTML-escaped before rendering to prevent XSS
• PDF uploads capped at 10 MB
• API key entered in-session only, never stored to disk