FastAPI server for document processing and RAG-based questioning.

1. python environment:

   python -m venv venv
   source venv/bin/activate  # on windows: venv\Scripts\activate

2. install dependencies:

   pip install -r requirements.txt

3. configure environment:

   Create a .env file:

   MISTRAL_API_KEY=your_mistral_api_key_here
   GROQ_API_KEY=your_groq_api_key_here

4. run the server:

   python main.py

• mistral api: you need a valid MISTRAL_API_KEY for embeddings.
• groq api: you need a valid GROQ_API_KEY for generation.
• models:
  • Embeddings: mistral-embed (handled via API)
  • Generation: llama-3.3-70b-versatile (handled via Groq API)

• POST /session: create a new session; returns session_id. Use this on app load.
• POST /upload: upload and index a PDF (query param: session_id). No disk write; state is in memory per session.
• POST /query: ask questions about the document for the given session (query params: question, session_id).
• GET /status: optional session_id for per-session index state.
• GET /health: health check.

backend/
├── main.py             # fastapi entry point & routes
├── session_store.py    # in-memory session state (chunks, FAISS index)
├── extraction.py       # pdf text extraction (stream-based)
├── chunking.py         # text segmentation strategies
├── embeddings.py       # embeddings & in-memory FAISS index build
├── generation.py       # rag logic & llm interaction
├── requirements.txt    # python dependencies
└── .env                # environment variables (local only)