AI Tutor Multi-Agent System

A multi-agent tutoring system built with FastAPI and Google's Gemini API, following Agent Development Kit (ADK) principles. The system uses a coordinator pattern where a main Tutor Agent routes student questions to specialized sub-agents.

System Architecture

Main Components

1. Tutor Agent (Coordinator)

   • Routes queries to appropriate specialist agents
   • Uses keyword matching and Gemini API for intent classification
   • Handles general educational queries

2. Math Agent (Maths Specialist)

   • Handles mathematics-related questions
   • Uses Calculator Tool for arithmetic operations
   • Provides step-by-step solutions and explanations

3. Physics Agent (Physics Specialist)

   • Handles physics-related questions
   • Uses Physics Constants Tool, Unit Converter Tool, and Physics Formula Tool
   • Provides comprehensive physics explanations with unit conversions and formulas

4. Chemistry Agent (Chemistry Specialist)

   • Handles chemistry-related questions
   • Uses Periodic Table Tool for element information lookup
   • Provides detailed chemical element data and explanations

Tools

• Calculator Tool: Performs basic arithmetic operations safely
• Physics Constants Tool: Provides access to fundamental physics constants
• Unit Converter Tool: Converts between different units of measurement
• Physics Formula Tool: Provides physics formulas and calculations
• Periodic Table Tool: Comprehensive periodic table data with element properties

Installation

1. Clone the repository:

git clone https://github.com/Mayankrai449/MultiAgentTutor.git
cd MultiAgentTutor

2. Install dependencies:

pip install -r requirements.txt

3. Set up environment variables:

cp .env.example .env
# Edit .env and add your GEMINI_API_KEY

4. Get your Gemini API key:
   • Go to Google AI Studio
   • Create a new API key
   • Add it to your .env file

Usage

Starting the Server

uvicorn main:app --reload

The server will start on http://localhost:8000

API Endpoints

POST /ask

Submit a question to the AI tutor system.

Request:

{
  "question": "What is 25 + 17?"
}

Response:

{
  "answer": "25 + 17 equals 42. This is a basic addition problem...",
  "agent_used": "Math Agent",
  "tools_used": ["Calculator Tool"]
}

GET /health

Check if the service is running.

Example Queries

Math Questions:

• "What is 25 + 17?"
• "Solve the equation 2x + 5 = 11"
• "Calculate 15 * 8"

Physics Questions:

• "What is Newton's second law?"
• "What is the speed of light?"
• "Convert 100 mph to km/h"
• "What is the formula for kinetic energy?"

Chemistry Questions:

• "Tell me about hydrogen"
• "What is the atomic mass of carbon?"
• "What elements are in group 18?"
• "Describe the electron configuration of oxygen"

General Questions:

• "How do I study effectively?"
• "What is the scientific method?"

Features

Multi-Agent Architecture

• Coordinator Pattern: Main agent routes queries to specialists
• Specialization: Each agent focuses on specific domains (Math, Physics, Chemistry)
• Modularity: Easy to add new agents and tools

Intelligent Routing

• Keyword-based classification for fast routing
• Gemini API fallback for complex classification
• Support for general educational queries

Comprehensive Tool Integration

• Calculator Tool: Safe mathematical expression evaluation
• Physics Constants Tool: Comprehensive physics constants database
• Unit Converter Tool: Convert between different measurement units
• Physics Formula Tool: Access to common physics formulas and calculations
• Periodic Table Tool: Complete periodic table with detailed element information including:
  • Atomic number and mass
  • Electron configuration
  • Element properties (group, period, state, category)
  • Descriptive information

Enhanced Response Format

• Detailed answers with step-by-step explanations
• Information about which agent handled the query
• List of tools used in generating the response

Deployment

Local Development

uvicorn main:app --reload --host 0.0.0.0 --port 8000

Deployed Link

Deployed Site

Configuration

Environment Variables

• GEMINI_API_KEY: Your Gemini API key (required)

Model Configuration

The system uses gemini-2.0-flash-001 by default. You can modify this in the agent files if needed.