Thin Multi-Agent System

A sophisticated multi-agent system demonstrating integration between different AI platforms and agent frameworks. This project showcases how to build a travel assistant ecosystem with specialized agents for different domains.

Project Intentions

This project demonstrates:

• Multi-Agent Architecture: Building a system with specialized agents that can work together to provide comprehensive travel assistance
• Platform Integration: Showcasing integration between Microsoft Copilot Studio and Azure AI Foundry agents
• Real-world Application: Creating practical travel and dining recommendation services
• Scalable Design: Implementing a flexible architecture that can accommodate additional agent types and capabilities

What's Included

🤖 Specialized Agents

Travel Explorer Agent (Copilot Studio)

• Recommends tourist attractions based on destination cities
• Categorizes attractions by type (religious sites, nightlife, family-friendly locations)
• Provides comprehensive travel guidance

Culinary Advisor Agent (Azure AI Foundry)

• Restaurant guide for cities worldwide
• Tailored dining recommendations based on user preferences
• Covers local delicacies, fine dining, budget options, and specific cuisines
• Provides information on popular dishes, ratings, and unique culinary experiences

🏗️ Architecture Components

Backend (/backend)

• FastAPI-based REST API with agent orchestration
• Intent routing system for multi-agent coordination
• Telemetry and tracing capabilities
• Chainlit UI integration for interactive chat
• Modular agent framework supporting multiple platforms

Frontend (/frontend)

• React-based visual interface with drag-and-drop functionality
• Interactive canvas for visualizing agent relationships
• Real-time chat panel integration
• Responsive design for various screen sizes

🔧 Key Features

• Intent Routing: Automatically routes user queries to the most appropriate agent
• Multi-Platform Support: Seamlessly integrates Copilot Studio and Azure AI Foundry agents
• Visual Interface: Interactive canvas for understanding agent relationships
• Real-time Chat: Live conversation interface with specialized agents
• Extensible Design: Easy to add new agents and capabilities
• Telemetry Integration: Built-in monitoring and tracing for performance insights

Technology Stack

• Backend: Python, FastAPI, Chainlit
• Frontend: React, TypeScript, React Flow
• AI Platforms: Microsoft Copilot Studio, Azure AI Foundry
• Infrastructure: Azure AI Services, REST APIs

Getting Started

For detailed setup and running instructions, please see SETUP.md.

Project Structure

thin-multi-agent/
├── backend/           # FastAPI backend with agent orchestration
│   ├── agents/        # Agent implementations
│   ├── api/          # REST API endpoints
│   ├── models/       # Data models and schemas
│   ├── orchestrator/ # Agent coordination logic
│   └── telemetry/    # Monitoring and tracing
├── frontend/         # React frontend application
│   ├── src/          # Source code
│   │   ├── components/ # UI components
│   │   └── context/   # React context providers
│   └── public/       # Static assets
└── SETUP.md         # Detailed setup instructions

Use Cases

• Travel Planning: Get comprehensive destination recommendations with categorized attractions
• Dining Discovery: Find the perfect restaurants based on your preferences and location
• Multi-Agent Workflows: Understand how different AI agents can collaborate effectively
• Platform Integration: Learn how to integrate multiple AI platforms in a single application

Contributing

This project serves as a demonstration of multi-agent system architecture and platform integration. Feel free to explore the code, experiment with different agent configurations, and extend the functionality.

License

This project is for demonstration purposes and showcases integration patterns for multi-agent AI systems.