PDF-RAG-AI: AI-Powered Document Question Answering end to end encrypted without anyone interruption (not even mine)

We not store your pdf at our side. It is chunked for safety

A modern web application that enables intelligent document analysis and question answering for PDF documents using RAG (Retrieval-Augmented Generation) technology.

🌟 Features

• PDF Upload & Processing: Upload PDFs for AI analysis and indexing
• Intelligent Question Answering: Ask natural language questions about your documents
• Context-Aware Responses: AI generates answers based specifically on your document content
• Secure Authentication: User authentication with Clerk
• Simple UI: Clean, responsive design with dark mode support

🏗️ Architecture

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│             │     │             │     │             │
│   Client    │────▶│   Server    │────▶│   Worker    │
│  (Next.js)  │     │  (Express)  │     │ (Background)│
│             │     │             │     │             │
└─────────────┘     └─────────────┘     └─────────────┘
       │                  │                   │
       │                  │                   │
       │                  ▼                   │
       │           ┌─────────────┐            │
       │           │             │            │
       └──────────▶│   Valkey    │◀──────────┘
                   │  (Redis)    │
                   │             │
                   └─────────────┘
                         │
                         │
                         ▼
                   ┌─────────────┐
                   │             │
                   │   Qdrant    │
                   │(Vector DB)  │
                   │             │
                   └─────────────┘

• Client: Next.js React application for the user interface
• Server: Express.js backend API handling PDF uploads and queries
• Worker: Background processing for PDF ingestion and vector embedding
• Valkey: Redis-compatible message broker for job queue
• Qdrant: Vector database for storing document embeddings

🔧 Tech Stack

Frontend (Client)

• Next.js 15.3 with App Router
• React 19
• TypeScript
• TailwindCSS 4
• ShadCN UI components
• Clerk Authentication
• React Toastify for notifications
• Embla Carousel for UI components

Backend (Server)

• Express.js 5
• LangChain for LLM operations
• Mistral AI for embeddings and chat completions
• BullMQ for background processing
• Multer for file uploads
• PM2 for process management

Databases & Services

• Qdrant vector database for semantic search
• Valkey Redis-compatible message broker
• Docker containerization

🚀 Getting Started

Prerequisites

• Node.js 20+
• Docker and Docker Compose

Running Locally

1. Clone the repository

git clone https://github.com/Sumit-Ks1/PDF-RAG-AI.git
cd PDF-RAG-AI

2. Set up environment variables:

   • Create server/.env with Mistral API key and connection details
   • Create client/.env.local with backend URL and Clerk keys

3. Start with Docker Compose:

docker-compose up --build

4. Access the application:
   • Frontend: http://localhost:3000
   • Backend API: http://localhost:8000

📁 Project Structure

├── client/              # Next.js frontend application
├── server/              # Express backend application
└── docker-compose.yml   # Docker Compose configuration

See individual README files in client/ and server/ directories for more details.

🧩 How It Works

1. Authentication: Users sign in through Clerk authentication
2. PDF Upload: Documents are uploaded to the server
3. Processing: Worker processes PDFs and creates vector embeddings
4. Storage: Embeddings stored in Qdrant vector database
5. Query: Users ask questions about their documents
6. Retrieval: System retrieves relevant document sections
7. Response: AI generates answers based on retrieved content

📄 License

This project is licensed under the ISC License.

🙏 Acknowledgments

• Mistral AI for the LLM model
• LangChain for the RAG framework
• Qdrant for vector search capabilities
• Clerk for authentication