Multi-Agent AI Tutor

A sophisticated AI tutoring system that implements advanced agent-based architecture for intelligent subject-specific tutoring. The system leverages Google's Gemini API and incorporates principles from Google's Agent Development Kit (ADK) to create a highly capable, context-aware tutoring experience.

Live Demo

Visit the live application at: https://multi-agent-ai-tutor.vercel.app/

Intelligent Agent Architecture

Our system implements a sophisticated multi-agent architecture inspired by modern AI agent frameworks, featuring:

Core Agent System

1. Tutor Agent (Orchestrator)

   • Implements advanced query classification and routing
   • Maintains conversation context and learning progression
   • Employs sophisticated prompt engineering for optimal responses
   • Uses reflection and self-improvement mechanisms

2. Specialized Subject Agents

   • Math Agent

     • Domain-specific knowledge representation
     • Advanced problem decomposition
     • Step-by-step solution generation
     • Mathematical reasoning validation

   • Physics Agent

     • Conceptual understanding verification
     • Physical law application
     • Unit conversion and dimensional analysis
     • Real-world application mapping

   • Chemistry Agent

     • Chemical reaction analysis
     • Molecular structure understanding
     • Stoichiometric calculations
     • Periodic trends application

Agent Capabilities

• Contextual Understanding

  • Recognizes user's learning level
  • Adapts explanations accordingly

• Tool Integration

  • Calculator: Advanced mathematical computations
  • Constant Lookup: Scientific constants and properties

• Learning Enhancement

  • Interactive problem-solving
  • Error analysis and correction

Technical Implementation

Modern Tech Stack

• Backend: FastAPI (Async-first, high-performance)
• Frontend: HTML, TailwindCSS (Responsive, modern UI)
• AI Engine: Google Gemini API (State-of-the-art LLM)
• Deployment: Vercel (Serverless, edge-optimized)

System Features

• Real-time query processing
• Asynchronous operation handling
• Robust error management
• Comprehensive logging and monitoring
• Scalable serverless architecture

Setup and Development

1. Environment Setup

   git clone <repository-url>
   cd ai-tutor
   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt

2. Configuration Create .env:

   GEMINI_API_KEY=your_api_key_here
   DEBUG=False
   ENVIRONMENT=development
   

3. Local Development

   uvicorn app.main:app --reload

API Architecture

• GET /: Interactive web interface
• POST /api/query: Intelligent query processing
• GET /api/health: System health monitoring
• GET /api/metrics: Performance metrics

Advanced Features

• Intelligent Routing

  • Context-aware query classification
  • Dynamic agent selection
  • Multi-agent collaboration when needed

• Learning Optimization

  • Adaptive difficulty levels
  • Personalized learning paths
  • Concept reinforcement
  • Progress tracking

• System Reliability

  • Graceful error handling
  • Fallback mechanisms
  • Performance monitoring
  • Automated recovery

Contributing

We welcome contributions! The system is designed to be extensible, allowing for:

• New subject agent integration
• Additional tool development
• UI/UX improvements
• Performance optimizations

License

This project is open-source and available under the MIT License.