Multi-Agent Trip Planning System

A sophisticated multi-agent workflow built with Google's Agent Development Kit (ADK) that demonstrates the power of specialized AI agents working together to solve complex travel planning tasks.

🏗️ Architecture

Instead of building one monolithic "super agent," this system employs a team of specialized AI agents, each expert in their domain:

Core Agents

All basic agents are consolidated in subagent.py:

1. FlightAgent - Flight booking specialist

   • Handles flight searches and bookings
   • Returns structured JSON with flight details
   • Makes reasonable assumptions when details are missing

2. HotelAgent - Hotel booking specialist

   • Manages hotel searches and reservations
   • Provides accommodation details in JSON format
   • Handles various room types and booking preferences

3. SightseeingAgent - Tourism specialist

   • Recommends top 2 attractions per destination
   • Provides timing and relevant details
   • Focuses on must-see locations

4. TripSummaryAgent - Summary compilation specialist

   • Compiles trip details into comprehensive itinerary
   • Creates structured travel summaries
   • Formats information for easy reading

Orchestration Agents

Each orchestration agent has its own folder with a dedicated agent.py:

5. SimpleAgent (simple/) - Basic trip coordinator

   • Simple sub-agent coordination pattern
   • Direct management of flight, hotel, and sightseeing agents
   • Perfect for straightforward trip planning

6. DispatcherAgent (dispatcher/) - Intelligent request router

   • Analyzes requests and routes to appropriate specialists
   • Uses agent tools for flexible coordination
   • Handles simple to complex multi-step requests

7. ParallelAgent (parallel/) - Efficiency optimizer

   • Runs flight and hotel agents in parallel for speed
   • Sequential execution: sightseeing → parallel(flight+hotel) → summary
   • Maximizes efficiency for independent flight and hotel tasks

8. SelfCriticAgent (self_critic/) - Quality assurance specialist

   • Same parallel execution as ParallelAgent (flight+hotel in parallel)
   • Adds quality control: trip summary reviewer and validator
   • Ensures output meets quality standards before delivery

Workflow Patterns

• Parallel Execution: Flight and hotel bookings run concurrently for efficiency
• Sequential Orchestration: Dependent tasks execute in logical order
• Feedback Loops: Built-in quality assurance and validation
• State Management: Agents communicate through shared session state

🚀 Quick Start

Prerequisites

• Python 3.8+
• Google API Key for Gemini models

Installation

1. Clone and navigate to the project:

   git clone <repository-url>
   cd adk_workflows

2. Create and activate virtual environment:

   # Create virtual environment
   python -m venv venv
   
   # Activate virtual environment
   # On macOS/Linux:
   source venv/bin/activate
   
   # On Windows:
   # venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment:

   # Copy environment template
   cp env.example .env
   
   # Edit .env and add your Google API key
   # Get your key from: https://aistudio.google.com/app/apikey

5. Launch the web interface:

   adk web

   This will open a web interface where you can choose and test any of the available agents.

📋 Configuration

Environment Variables (.env)

# Required
GOOGLE_API_KEY=your_google_api_key_here
MODEL_NAME=gemini-2.0-flash

# Optional
ENVIRONMENT=development
LOG_LEVEL=INFO

🏃‍♂️ Usage

Once you've completed the installation steps, simply run:

adk web

This opens a web interface where you can:

• Select any of the 4 orchestration agents (Simple, Dispatcher, Parallel, Self-Critic)
• Test different types of trip planning requests
• See how each agent handles various scenarios

Example Requests You Can Test

• Simple: "Find me a flight to Paris"
• Complex: "Book a flight to Paris and find a hotel near the Eiffel Tower"
• Comprehensive: "Plan a 3-day trip to Tokyo with flights, accommodation, and sightseeing"

🎯 Agent Details

FlightAgent

• Purpose: Specialized flight booking and information
• Input: Flight preferences, dates, destinations
• Output: JSON with flight details, prices, booking status
• Features: Intelligent assumptions for missing details

HotelAgent

• Purpose: Hotel booking and accommodation management
• Input: Location, dates, room preferences
• Output: JSON with hotel details, pricing, availability
• Features: Room type optimization, location-based suggestions

SightseeingAgent

• Purpose: Tourism recommendations and itinerary planning
• Input: Destination, interests, duration
• Output: JSON with top 2 attractions, timings, details
• Features: Curated recommendations, practical timing info

TripSummaryAgent

• Purpose: Quality assurance and trip compilation
• Components:
  • TripSummaryAgent: Compiles comprehensive itinerary
  • TripSummaryReviewer: Quality check and validation
  • ValidateTripSummary: Final approval and feedback
• Output: Validated, complete travel itinerary

🔄 Workflow Options

Simple Workflow

TripPlanner (root_agent) → Coordinates FlightAgent + HotelAgent + SightseeingAgent

Dispatcher Workflow

DispatcherAgent → Analyzes request → Routes to appropriate tools → Compiles response

Parallel Workflow

SightseeingAgent → FlightAgent + HotelAgent (parallel) → TripSummaryAgent

Self-Critic Workflow

SightseeingAgent → FlightAgent + HotelAgent (parallel) → TripSummaryAgent → Reviewer → Validator

🧪 Development

Project Structure

adk_workflows/
├── subagent.py            # All core agents (flight, hotel, sightseeing, trip_summary)
├── simple/
│   └── agent.py           # Basic trip coordinator
├── dispatcher/
│   └── agent.py           # Intelligent request router
├── parallel/
│   └── agent.py           # Parallel execution optimizer
├── self_critic/
│   └── agent.py           # Quality assurance workflow
├── requirements.txt       # Dependencies
├── env.example           # Environment template
├── README.md             # Documentation
└── SETUP_INSTRUCTIONS.md # Setup guide

Adding New Agents

Core Agents:

1. Add your new core agent to subagent.py
2. Import and use in orchestration agents as needed

Orchestration Agents:

1. Create a new folder: new_orchestrator/
2. Add agent.py with your orchestration logic
3. Import core agents from subagent.py

🎯 Benefits of Multi-Agent Architecture

1. Specialization: Each agent excels in its specific domain
2. Scalability: Easy to add new agents or modify existing ones
3. Maintainability: Clear separation of concerns
4. Efficiency: Parallel execution for independent tasks
5. Quality: Built-in review and validation processes
6. Flexibility: Modular design allows easy customization

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Add your specialized agent or enhancement
4. Update documentation
5. Submit a pull request

📜 License

[Add your license information here]

🆘 Support

For issues and questions:

• Open an issue in the repository
• Check the Google ADK documentation
• Review the agent implementation examples

---

Built with Google Agent Development Kit (ADK) - Empowering intelligent multi-agent workflows with Gemini.