TravelBot AI

An intelligent travel companion powered by LangGraph, RAG (Retrieval-Augmented Generation), and the TravelPlanner dataset. TravelBot AI helps users plan their trips by providing information about destinations, weather, and some useful calculations through an interactive chat interface.

Features

• 🔍 Knowledge Search: Access to comprehensive travel information from the TravelPlanner dataset
• 🌤️ Weather Information: Real-time weather data for any destination
• 🧮 Travel Calculator: Calculate costs, distances, and other travel-related math
• 🤖 Conversational AI: Powered by Hugging Face transformers with LangGraph workflow
• 📊 Vector Search: FAISS-powered semantic search across travel data
• 🎯 Interactive UI: Clean Streamlit interface with chat history and tool execution tracking

Architecture

• LangGraph: Orchestrates the AI agent workflow with conditional routing
• RAG Pipeline: Combines retrieval from vector database with generative AI
• FAISS Vector Store: Efficient similarity search across travel documents
• Hugging Face Models: Local model execution with transformer support
• Streamlit: Modern web interface for user interaction

Prerequisites

• GPU recommended for optimal performance
• API keys for enhanced functionality (required)

Installation

1. Clone the repository

   git clone <https://github.com/37nomad/AI-Travel-Agent.git>
   cd AI-Travel-Agent

2. Install dependencies

   pip install -r requirements.txt

3. Set up environment variables

   Create a .env file in the project root:

   HUGGINGFACE_TOKEN=your_huggingface_token_here
   OPENWEATHER_API_KEY=your_openweather_api_key_here

   Getting API Keys:

   • Hugging Face Token: Sign up at huggingface.co, go to Settings → Access Tokens
   • OpenWeather API Key: Register at openweathermap.org for free API access

4. Run the setup script

   python setup_script.py

   This will:

   • Create necessary directories
   • Download and process the TravelPlanner dataset (500-1000 entries)
   • Build the FAISS vector index
   • Test the installation

Quick Start

1. Launch the application

   streamlit run main.py

2. Open your browser and navigate to the displayed URL (typically http://localhost:8501)

Usage Examples

Try these example queries in the chat interface:

• "Find hotels in Paris for 2 nights"
• "What's the weather like in Tokyo?"
• "Calculate the total cost: flight ₹15000 + hotel ₹8000"
• "Tell me about restaurants in New York"
• "Find attractions in London"

Configuration

Model Selection

The default model is microsoft/DialoGPT-medium. For better performance, consider upgrading to a paid/premium model by modifying the model_name parameter in agent.py:

agent = TravelAgent(model_name="your-preferred-model")

Dataset Limit

Adjust the dataset processing limit in setup_script.py or when running directly:

# Process 500 entries (recommended for testing)
processor.process_and_save_dataset(limit=500)

# Process 1000 entries (maximum, better coverage)
processor.process_and_save_dataset(limit=1000)

Tools Available

1. Knowledge Search

Search the travel database for information about:

• Hotels and accommodations
• Restaurants and dining
• Tourist attractions
• Destination information

2. Weather Information

Get current weather data for any city worldwide using OpenWeather API.

3. Calculator

Perform travel-related calculations:

• Cost estimation
• Distance calculations
• Currency conversion
• General arithmetic

Performance Tips

• GPU Usage: The application will automatically use GPU if available via device_map="auto"
• Memory: Vector index loading requires sufficient RAM
• First Run: Initial setup takes longer due to model and dataset downloads
• Batch Processing: Large dataset limits may require more processing time

What more could have been done

• Additional travel APIs integration
• Enhanced Tool Scraping Capability Or Using Enhanced features for smooth tool extraction
• Multi-language support
• Performance optimizations

Acknowledgments

• TravelPlanner Dataset: OSU NLP Lab for the comprehensive travel dataset

---