Intelligent Document Query Engine

A cloud-native, LLM-powered query-retrieval system designed to perform contextual analysis on large, unstructured documents. Built for scalable, production-ready deployments.

🚀 Features

• Retrieval-Augmented Generation (RAG) pipeline using:

  • faiss-cpu for efficient vector indexing
  • sentence-transformers CrossEncoder for high-precision reranking

• Llama 3 LLM integration via the Groq API for final answer synthesis

• Fine-tuned role-based system prompt for domain-specific query responses

• Lazy-loading pattern for large AI models to optimize memory and performance

• FastAPI backend with REST endpoints for document ingestion and query answering

• Docker-based deployment for reproducibility and scalability

🌐 Live Demo

The application is deployed on Railway and available here: API Documentation (Swagger UI)

📂 Project Structure

├── main.py                 # FastAPI entry point
├── requirements.txt        # Dependencies
├── Dockerfile              # Container build setup
├── start.py / start.sh     # Application startup scripts
├── test_*.py               # API test scripts
└── deploy.sh / deploy.bat  # Deployment scripts

⚙️ Installation & Setup

1. Clone the repository

   git clone https://github.com/yourusername/intelligent-document-query-engine.git
   cd intelligent-document-query-engine

2. Install dependencies

   pip install -r requirements.txt

3. Set environment variables Create a .env file with:

   GROQ_API_KEY=your_groq_api_key
   

4. Run locally

   uvicorn main:app --reload

☁️ Deployment

This project supports Railway deployment (Docker-based):

railway up

or locally via Docker:

docker build -t doc-query-engine .
docker run -p 8000:8000 doc-query-engine

📈 Performance Optimizations

• Lazy loading of embedding and reranker models
• FAISS in-memory index creation for efficient retrieval
• Batched query processing for speed
• Role-based prompts to reduce token usage

🧠 Tech Stack

• Backend: Python, FastAPI
• Vector Indexing: FAISS
• LLM API: Groq (Llama 3)
• Reranking: SentenceTransformers CrossEncoder
• Deployment: Docker, Railway