LLM Tour Recommendation

An intelligent travel planning system that generates optimized daily schedules and recommendations for travelers in Busan using LangChain and multiple AI agents.

🌟 Features

• Dual-Agent System:
  • Schedule Generator: Crafts detailed daily itineraries based on attraction visit times.
  • Recommendation Generator: Offers contextual travel tips and weather precautions.
• Weather-Aware Planning: Adapts to temperature, air quality, and weather conditions.
• Dynamic Scheduling: Adjusts plans for optimal travel experiences.
• GPU Acceleration: Supports M1 Macs with Metal Performance Shaders (MPS).

📋 Prerequisites

• Python 3.7+
• PyTorch
• LangChain
• Ollama

🚀 Installation

1. Clone the repository:

   git clone https://github.com/Valiev-Koyiljon/LLM_TOUR_RECOMMENDATION.git
   cd LLM_TOUR_RECOMMENDATION

2. Install required packages:

   pip install torch langchain ollama

3. Install Ollama and download the llama3.2 model:

   # Install Ollama (varies by OS)
   curl https://ollama.ai/install.sh | sh
   
   # Pull the model
   ollama pull llama3.2

💻 Usage

The system processes weather data and attraction information to create personalized travel plans. Here's an example:

# Sample data
data = [{
    "date": "2024-10-05",
    "temp_high": 30,
    "temp_low": 14,
    "weather": "Windy",
    "air_quality": "Hazardous",
    "attractions": [
        {"name": "Haeundae Beach", "recommended_time": 2.98},
        {"name": "Gwangalli Beach", "recommended_time": 3.18},
        {"name": "Taejongdae", "recommended_time": 3.63},
        {"name": "Beomeosa Temple", "recommended_time": 2.4},
        {"name": "Gamcheon Culture Village", "recommended_time": 1.0}
    ]
}]

# Initialize agent and process the data
agent = initialize_agent(tools, llm, agent=AgentType.OPENAI_FUNCTIONS, verbose=True)
for entry in data:
    result = process_prompt(agent, entry["date"], entry["temp_high"], 
                            entry["temp_low"], entry["weather"], 
                            entry["air_quality"], entry["attractions"])
    print(result)

🏗️ System Architecture

The system leverages two key AI agents:

1. Schedule Generator

   • Designs efficient daily schedules.
   • Considers weather and attraction visit times.
   • Optimizes visit order for convenience.

2. Recommendation Generator

   • Provides weather-specific advice.
   • Recommends precautions based on air quality.
   • Shares tips to enhance travel enjoyment.

📝 Input Data Format

Input data should follow this JSON structure:

{
    "date": "YYYY-MM-DD",
    "temp_high": float,
    "temp_low": float,
    "weather": string,
    "air_quality": string,
    "attractions": [
        {
            "name": string,
            "recommended_time": float
        }
    ]
}

🔧 Configuration

• GPU acceleration is enabled for M1 Macs via MPS.
• The system utilizes the llama3.2 model through Ollama.
• Agent verbosity is adjustable during initialization.