🧠 PDF Question-Answering API with Mistral

This project is a local API that allows you to upload a PDF, automatically index its content using FAISS and SentenceTransformers, and ask semantic questions answered by a local LLM (Mistral 7B Instruct via llama-cpp-python).

---

Features

• Upload and validate PDF files
• Extract and chunk text from PDF
• Create vector index (FAISS + all-MiniLM-L6-v2)
• Ask semantic questions using a local Mistral model
• No internet required — works fully offline after model download

---

Project Structure

pdf_qa_api/
├── app/
│ ├── main.py
│ ├── pdf_utils.py
│ ├── vector_store.py
│ ├── llm_interface.py
│ └── config.py
├── models/
├── uploads/
├── requirements.txt
└── README.md

---

Quickstart

1. Clone this repo
2. Download a .gguf model (e.g. Mistral 7B Instruct) into models/
3. Install dependencies:

pip install -r requirements.txt

4. Run the API:

uvicorn app.main:app --reload

5. Open the docs: http://localhost:8000/docs

Endpoints

• POST /upload_pdf: Upload a PDF (max 10MB) → triggers indexing
• GET /answer?question=...: Ask a semantic question about the PDF

Dependencies

• FastAPI
• PyMuPDF
• sentence-transformers
• faiss-cpu
• llama-cpp-python

Security

• Validates MIME type and size
• Refuses non-PDF files

License

This project is licensed under the MIT License - see the LICENSE file for details.