📚 PDF Chat with Vector Search

A full-stack application that allows users to upload PDFs, automatically parse and embed them into a PostgreSQL vector database, and chat with an AI assistant about the PDF contents using semantic search.

✨ Features

• 📄 PDF Upload & Processing: Upload PDFs with automatic text extraction and chunking
• 🤖 AI-Powered Chat: Chat with an intelligent assistant about your PDF contents
• 🔍 Semantic Search: Find relevant information using vector similarity search
• 💬 Chat History: Persistent conversation history for each PDF
• ⚡ Streaming Responses: Real-time AI responses for better user experience
• 📊 Performance Monitoring: Built-in timing and performance logging

🏗️ Architecture

• Backend: Node.js + Express + TypeScript + Prisma
• Frontend: React 18 + TypeScript + Redux Toolkit + Tailwind CSS
• Database: PostgreSQL with pgvector extension for vector operations
• AI Services: OpenAI API for embeddings and chat completions
• Vector Search: HNSW indexing for efficient similarity search

🚀 Quick Start

Prerequisites

• Node.js 18+
• PostgreSQL 13+ with pgvector extension
• OpenAI API key

1. Clone the Repository

git clone https://github.com/codefa/pdf-chat-vector.git
cd pdf-chat-vector

2. Set Up Database

Install PostgreSQL and the pgvector extension:

# Ubuntu/Debian
sudo apt-get install postgresql postgresql-contrib
sudo apt-get install postgresql-13-pgvector

# macOS with Homebrew
brew install postgresql
brew install pgvector

Create a database and enable the pgvector extension:

CREATE DATABASE pdf_chat;
\c pdf_chat
CREATE EXTENSION vector;

3. Environment Setup

Create a .env file backend/:

DATABASE_URL="postgresql://username:password@localhost:5432/pdf_chat"
OPENAI_API_KEY="your-openai-api-key-here"
PORT=5000

4. Install Dependencies

# Backend
cd backend
npm install

# Frontend (in another terminal)
cd frontend
npm install

5. Database Migration

cd backend
npx prisma migrate dev
npx prisma generate

6. Start the Application

# Backend (Terminal 1)
cd backend
npm run dev

# Frontend (Terminal 2)
cd frontend
npm run dev

The application will be available at:

• Frontend: http://localhost:5173
• Backend API: http://localhost:5000

📖 Usage

1. Upload a PDF

• Navigate to the home page
• Enter a title for your PDF
• Select a PDF file
• Click "Upload" to process the document

2. Chat with Your PDF

• Go to the PDFs list page
• Click "Chat" on any uploaded PDF
• Ask questions about the document content
• Get AI-powered responses based on semantic search

3. Manage PDFs

• View all uploaded PDFs
• Delete PDFs and related data
• Access chat history for each document

🗄️ API Endpoints

PDF Routes

• POST /api/pdf/upload - Upload and process PDF
• GET /api/pdf/list - List all PDFs
• DELETE /api/pdf/:id - Delete PDF and related data

Chat Routes

• POST /api/chat/:pdfId - Start/continue chat with streaming
• GET /api/chat/history/:chatId - Get chat history
• GET /api/chat/bench/:pdfId - Performance benchmarking

🗄️ Database Schema

-- PDF metadata
PDF (id, title, filename, createdAt)

-- Vector embeddings for text chunks
Embeddings (id, pdfId, chunk, embedding)

-- Chat sessions
Chat (id, pdfId, createdAt)

-- Conversation messages
Message (id, chatId, role, content, createdAt)

🛠️ Technology Stack

Backend

• Runtime: Node.js with TypeScript
• Framework: Express.js
• Database: PostgreSQL + pgvector
• ORM: Prisma
• AI: OpenAI API
• File Processing: pdf-parse

Frontend

• Framework: React 18 with TypeScript
• State Management: Redux Toolkit
• Routing: React Router DOM
• Styling: Tailwind CSS
• Build Tool: Vite

⚙️ Configuration

Vector Search Parameters

The application uses optimized HNSW indexing with configurable parameters:

// In embeddings service
await tx.$executeRawUnsafe('SET LOCAL diskann.query_rescore = 400')
await tx.$executeRawUnsafe('SET LOCAL diskann.query_search_list_size = 100')
await tx.$executeRawUnsafe('SET LOCAL hnsw.ef_search = 50')

AI Model Configuration

• Embeddings: text-embedding-3-small
• Chat: gpt-4o-mini
• Temperature: 0.2
• Max Tokens: 256

📊 Performance Features

• Chunking Strategy: ~1000 character chunks for optimal embedding
• Top-K Retrieval: Retrieves top 3 most relevant chunks
• Streaming Responses: Real-time AI responses
• Performance Logging: Built-in timing metrics

🚀 Deployment

Docker (Recommended)

# Backend Dockerfile
FROM node:18-alpine
WORKDIR /app
COPY package*.json ./
RUN npm ci --only=production
COPY . .
EXPOSE 5000
CMD ["npm", "start"]

Environment Variables for Production

NODE_ENV=production
DATABASE_URL="your-production-database-url"
OPENAI_API_KEY="your-production-openai-key"
PORT=5000

🤝 Contributing

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

Development Guidelines

• Follow TypeScript best practices
• Maintain RESTful API design principles
• Add tests for new features
• Update documentation as needed

📝 License

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

�� Acknowledgments

• OpenAI for AI services
• pgvector for PostgreSQL vector operations
• Prisma for database management
• Tailwind CSS for styling

📞 Support

If you encounter any issues or have questions:

1. Check the Issues page
2. Create a new issue with detailed information
3. Include your environment details and error logs

---

Made with ❤️ for developers who love AI and vector search!