FlyGPT - AI Agent for Smart Flight Search

Inspired by Zomunk and the problems I faced while searching for the best flight deals, FlyGPT is an AI Agent that searches on multiple dates or a month and returns the list of flights which has the best duration and price.

🎯 The Problem

Finding the best flight deals is a nightmare:

• Manual Date Checking: Searching day by day across a month
• Time Consuming: Hours spent comparing prices and durations
• Missed Deals: Best prices often hidden on specific dates
• Complex Queries: Natural language requests need manual interpretation

🚀 The Solution

FlyGPT is a FASTAPI wrapper on top of fast-flights with intelligent AI agents built using the Agno Framework. It automatically:

• Expands Date Ranges: "June 2025" → searches all 30 days
• Natural Language Processing: Understands conversational queries
• Parallel Search: Searches multiple dates simultaneously
• Smart Ranking: Returns top deals based on price, duration, and stops

🤖 Agent Architecture

The system uses two specialized AI agents:

Agent 1: Query Transformer

• Purpose: Converts natural language to structured search parameters
• Input: "Find me best flights from Bangalore to Ahmedabad in June"
• Output: Structured query with airports, dates, passengers, preferences
• Technology: Claude 3.5 Sonnet via Agno Framework

Agent 2: Flight Analyzer

• Purpose: Analyzes Google Flights responses and ranks best deals
• Input: Raw flight data from fast-flights across multiple dates
• Output: Top 10 ranked flights with price, duration, and stop analysis
• Technology: Claude 3.5 Sonnet via Agno Framework

🔄 Agent Flow Chart

The complete flow from natural language query to ranked flight results

🏗️ Technical Stack

• Backend: FastAPI + Python
• AI Framework: Agno Framework
• LLM: Claude 3.5 Sonnet (Anthropic)
• Flight Data: fast-flights (Google Flights wrapper)
• Data Models: Pydantic for validation
• API: RESTful endpoints with search ID tracking

🚀 Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Set Environment Variables

# Create .env file
ANTHROPIC_API_KEY=your_claude_api_key_here
FLIGHTS_API_HOST=0.0.0.0
FLIGHTS_API_PORT=3001

3. Run the API

python3 api/app.py

🔧 API Usage

Search Flights Endpoint

POST /api/v1/searchFlights/{search_id}

Request Body:

{
  "user_query": "Find me best flights from Bangalore to Ahmedabad in June 2025"
}

Response:

{
  "search_id": "search_123456",
  "summary": "Best flight options for 1 adult from Bangalore to Ahmedabad in June 2025",
  "flights": [
    {
      "airline": "IndiGo",
      "departure": "2025-06-15T08:30:00",
      "arrival": "2025-06-15T10:45:00",
      "price": "₹3,500",
      "duration": "2h 15m",
      "stops": 0
    }
  ]
}

🌟 Key Features

• 🎯 Multi-Date Search: Automatically expands "June" to all June dates
• 🚀 Parallel Processing: Searches 30 dates simultaneously vs. sequentially
• 🤖 AI-Powered: Natural language understanding and intelligent ranking
• 📊 Smart Analysis: Multi-criteria optimization (price, duration, stops)
• 🔍 Search Tracking: Unique search ID for debugging and support
• 🌐 Google Flights: Leverages Google's comprehensive flight data via fast-flights

🏆 Why This Architecture?

• Separation of Concerns: Each agent has a focused responsibility
• Scalability: Easy to add new agents or modify existing ones
• Maintainability: Clear data contracts between agents
• Reliability: Robust error handling and validation
• User Experience: Natural language to structured results

📁 Project Structure

final_fly_gpt/
├── api/
│   └── app.py              # FastAPI application
├── src/
│   ├── agno_agent.py       # AI agent definitions
│   ├── core.py             # Flight search engine
│   ├── main.py             # Main orchestration
│   ├── json_schema.py      # Data models
│   └── logger.py           # Logging utilities
├── tests/                  # Test files
├── requirements.txt        # Python dependencies
└── README.md              # This file

🤝 Contributing

This project is open source and welcomes contributions! Please read our Contributing Guidelines for details.

📄 License

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

---

All hail Google Flights and fast-flights! 🛫

Built with ❤️ using Agno Framework and Claude 3.5 Sonnet

Inspired by the amazing flight deals from Zomunk