AI Chatbot with DeepSeek R1 and Local RAG- 🤖 Dual-Mode Chat: Switches between campus-specific and general knowledge

A full-stack, Retrieval-Augmented Generation (RAG) powered chatbot application for campus information. This AI system answers campus-specific questions using a local Q&A dataset and the deepseek-r1:1.5b model via Ollama. It also works as a general-purpose chatbot for unrelated queries.

---

📑 Table of Contents

• ✨ Features
• 🏗️ Project Structure
• ⚙️ Installation
  • Install Ollama
  • Pull AI Models
  • Install Python Dependencies
• 🚀 Running the App
• 🔄 How RAG Works
• 🛠️ Customization
• 🧪 Testing
• ❓ FAQ
• 🤝 Contributing
• 📄 License

---

✨ Features

• 🎯 Accurate Campus Information: Uses srec_qa.json with RAG for precise answers about your campus
• � Dual-Mode Chat: Switches between campus-specific and general knowledge
• ⚡ Local & Private: Runs completely offline with Ollama models
• 🎨 Modern Web Interface: Clean design with real-time chat updates
• 🔄 RAG Integration: Smart context retrieval for accurate campus-related responses
• 🛠️ Modular Architecture: Easy to customize and extend with utils/ helpers
• 🧪 Test Suite: Includes test_chatbot.py for reliability

---

🏗️ Project Structure

AI-ChatBot/
├── app.py                    # Flask backend with RAG integration
├── requirements.txt          # Python dependencies
├── test_chatbot.py          # Test suite for chatbot functionality
├── Campus_qa.json              # Q&A dataset for RAG
├── frontend/                 # Frontend assets
│   ├── index.html           # Main UI template
│   ├── app.js               # Frontend logic and API calls
│   ├── style.css            # UI styling
│   └── brain.png            # Logo asset
└── utils/                   # Utility modules
    └── time_helper.py       # Time-related utilities

---

⚙️ Installation

Install Ollama

Choose your platform:

📦 Windows

1. Download from ollama.com/download
2. Run the installer
3. Start Ollama from Start Menu

🐧 Linux

curl -fsSL https://ollama.com/install.sh | sh

🍎 macOS

curl -fsSL https://ollama.com/install.sh | sh

Pull AI Models

# Pull required models
ollama pull deepseek-r1:1.5b
ollama pull mxbai-embed-large

# Verify installation
ollama list

Install Python Dependencies

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # Linux/macOS
venv\Scripts\activate   # Windows

# Install dependencies
pip install -r requirements.txt

---

🚀 Running the App

1. Start Ollama Server

   ollama serve

2. Launch Backend

   python app.py

3. Access the UI

   • Open http://localhost:5000
   • Start chatting! 💬

---

🔄 How Retrieval-Augmented Generation (RAG) Works

1. Question Analysis 🔍

   • Detects campus-specific queries
   • Routes general queries to standard chat

2. Vector Search 📊

   • Embeds questions using mxbai-embed-large
   • Performs similarity search in campus Q&A

3. Context Injection 🎯

   • Enriches prompts with relevant campus info
   • Ensures accurate, grounded responses

4. Response Generation ✨

   • Uses Gemma2:2b for natural language
   • Maintains conversation context

---

🛠️ Customization

Adding Campus Knowledge

// campus_qa.json
{
  "questions": [
    {
      "q": "What are the campus timings?",
      "a": "Your custom answer here"
    }
  ]
}

Model Configuration

• Swap models in app.py:

CHAT_MODEL = "deepseek-r1:1.5b"  # Or any Ollama model
EMBED_MODEL = "mxbai-embed-large"

---

🧪 Testing

Run the test suite:

python test_chatbot.py

API Testing:

# Health check
curl http://localhost:5000/health

# Chat endpoint
curl -X POST http://localhost:5000/chat/simple \
     -H "Content-Type: application/json" \
     -d '{"message": "What are the campus timings?"}'

---

❓ FAQ

Q: How is data privacy maintained?

All processing happens locally. No data leaves your system.

Q: Can I extend the campus knowledge?

Yes! Add Q&A pairs to campus_qa.json.

Q: What's the recommended hardware?

- RAM: 8GB minimum, 16GB recommended - GPU: Optional, CPU works fine - Storage: 10GB for models

---

🤝 Contributing

We welcome contributions! See our Contributing Guide for details.

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

---

📄 License

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

---

Built with ❤️ using Gemma2 and Flask