PDF Question-Answering API

This project is a FastAPI-based application that downloads PDFs, extracts text, builds a FAISS vector index using sentence-transformers/all-MiniLM-L6-v2, and answers questions using the Mistral model via Ollama. It processes various documents (e.g., Indian Constitution, insurance policies, technical manuals) and handles queries through a REST API endpoint.

Features

• PDF Processing: Downloads and extracts text from PDFs.
• Vector Indexing: Uses FAISS and HuggingFace embeddings for efficient document retrieval.
• Question Answering: Leverages the Mistral model (Ollama) to answer questions based on PDF content.
• Concurrency Control: Limits parallel requests to prevent server overload.
• Retry Logic: Handles transient failures (e.g., timeouts) with retries.

Prerequisites

• OS: Windows (tested), Linux, or macOS
• Python: 3.10
• GPU: NVIDIA GPU with CUDA for optimal performance (CPU fallback supported)
• Ollama: Installed and running with the Mistral model
• Dependencies: Listed in requirements.txt

Setup Instructions

Clone the Repository

git clone https://github.com/abhinai2244/LLM-Powered-Intelligent-Query-Retrieval-System.git
cd LLM-Powered-Intelligent-Query-Retrieval-System

Set Up a Virtual Environment

python -m venv venv
# Windows
.\\venv\\Scripts\\activate
# Linux/macOS
# source venv/bin/activate

Install Dependencies

pip install -r requirements.txt

Requirements include:

• fastapi
• uvicorn
• llama-index
• llama-index-embeddings-huggingface
• llama-index-llms-ollama
• llama-index-vector-stores-faiss
• requests
• sentence-transformers
• faiss-gpu (or faiss-cpu if no GPU)
• tenacity
• fastapi-concurrency

Install and Configure Ollama

1. Download and install Ollama: ollama.com

2. Start the Ollama server:

   ollama serve

3. Pull the Mistral model:

   ollama pull mistral

4. Verify Ollama is running:

   curl http://localhost:11434

Verify GPU Support

1. Check GPU availability:

   nvidia-smi

2. Ensure faiss-gpu is installed for GPU acceleration. Use faiss-cpu if no GPU is available:

   pip install faiss-cpu

Usage

Run the FastAPI Server

cd <your-repo>
uvicorn new:app --host 127.0.0.1 --port 8000 --workers 1

The server will be available at http://127.0.0.1:8000.

Send a Request

Use curl, Postman, or a script to send a POST request to /api/v1/hackrx/run. Example:

curl -X POST http://127.0.0.1:8000/api/v1/hackrx/run -H "Content-Type: application/json" -d '{
  "documents": "https://hackrx.blob.core.windows.net/assets/indian_constitution.pdf?sv=2023-01-03&st=2025-07-28T06%3A42%3A00Z&se=2026-11-29T06%3A42%3A00Z&sr=b&sp=r&sig=5Gs%2FOXqP3zY00lgciu4BZjDV5QjTDIx7fgnfdz6Pu24%3D",
  "questions": [
    "What is the official name of India according to Article 1 of the Constitution?",
    "What is abolished by Article 17 of the Constitution?"
  ]
}'

Response:

{
  "answers": [
    "The official name of India, according to Article 1 of the Constitution, is 'India, that is Bharat'.",
    "Article 17 of the Constitution abolishes 'untouchability' and forbids its practice in any form."
  ]
}

Supported PDFs and Questions

The API has been tested with:

• Indian Constitution: Legal questions (e.g., rights, articles).
• Arogya Sanjeevani Policy: Insurance claims (e.g., root canal, IVF coverage).
• Super Splendor Manual: Technical questions (e.g., spark plug gap).
• Family Medicare Policy: Health insurance queries.
• Principia Newton: Physics questions (e.g., laws of motion).
• UNI Group Health Insurance: Coverage details.

Ensure the PDF URL matches the question context to avoid irrelevant answers.

Troubleshooting

Ollama Timeout (httpx.ReadTimeout):

• Increase timeout in new.py:

  Settings.llm = Ollama(model="mistral", request_timeout=1200.0)

• Restart Ollama:

  # Windows
  taskkill /IM ollama.exe /F
  ollama serve

• Check GPU usage:

  nvidia-smi

• Use a smaller model:

  ollama run mistral:7b

DNS Issues:

If pulling Mistral fails (no such host):

nslookup dd20bb891979d25aebc8bec07b2b3bbc.r2.cloudflarestorage.com
netsh interface ip set dns name="Wi-Fi" source=static address=8.8.8.8
ipconfig /flushdns
ollama pull mistral

Concurrent Request Overload:

The server limits concurrency to 1 request. Ask clients to avoid parallel requests or increase max_concurrent in fastapi-concurrency.

Port Conflict:

If port 11434 (Ollama) or 8000 (FastAPI) is in use:

netstat -aon | findstr :11434
taskkill /PID <PID> /F

Dependency Issues:

Ensure compatible versions:

pip install langchain==0.2.16 langchain-community==0.2.16 pydantic==2.8.2

Project Structure

<your-repo>/
├── new.py              # FastAPI application
├── requirements.txt    # Dependencies
├── docs/               # Directory for downloaded PDFs
├── index_<hash>/       # Cached FAISS indices
└── README.md           # This file

Contributing

1. Fork the repository.
2. Create a feature branch (git checkout -b feature/your-feature).
3. Commit changes (git commit -m "Add your feature").
4. Push to the branch (git push origin feature/your-feature).
5. Open a pull request.

License

MIT License. See LICENSE for details.

Contact

For issues or questions, open a GitHub issue or contact <your-email>.