An advanced AI-powered pirate treasure hunting game where three specialized AI agents work together to navigate treacherous waters, collect treasures, and battle enemies using LangGraph and local LLM integration.

Navigate a pirate ship across a 30x30 grid map using three AI agents that collaborate to make strategic decisions:

• 🧭 Navigator: Scans the environment and provides reconnaissance reports
• ⚔️ Cannoneer: Handles combat operations and threat assessment
• 👨‍✈️ Captain: Makes final strategic decisions and coordinates the crew

• LangGraph Integration: Advanced AI workflow with specialized agent roles
• Local LLM Support: Uses Ollama for private, local AI processing
• Custom System Prompts: Editable prompts via web interface for each agent
• Real-time Decision Making: Agents analyze game state and make tactical decisions

• Responsive 3-Column Layout: System prompts, game map, and agent outputs
• Material Icons: Professional UI with Google Material Design icons
• Real-time Updates: Live agent responses and tool outputs via polling
• Chrome DevTools Integration: Live editing capabilities with workspace mapping

• Strategic Movement: Up to 3-tile movement per turn with path validation
• Resource Management: Limited cannonballs with treasure collection rewards
• Dynamic Enemies: Moving enemies and monsters with AI behavior
• Multiple Objectives: Collect treasures while avoiding or defeating threats

• Python 3.9+
• Ollama installed and running locally
• Chrome browser (recommended for DevTools integration)

1. Clone the repository

   git clone https://github.com/bradmonk/pirates.git
   cd pirates

2. Set up Python environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Install an Ollama model

   ollama pull llama3.1:latest

5. Start the game

   ./restart.sh

6. Open the web interface

   • Navigate to http://localhost:8000 in Chrome
   • Select an AI model from the dropdown
   • Click "Start Game" and watch the AI agents play!

• Use the left panel to modify each agent's behavior
• Real-time prompt updates without restarting the game
• Experiment with different strategic approaches

pirates/
├── index.html              # Main web interface
├── styles.css              # Separated CSS styling
├── web_gui.py              # HTTP server and web interface logic
├── ai_agents.py            # LangGraph AI agents implementation
├── game_state.py           # Game mechanics and state management
├── game_tools.py           # Agent tools for game interaction
├── pirate_game.py          # Main game coordination
├── map.csv                 # Game map definition
├── requirements.txt        # Python dependencies
├── restart.sh              # Development startup script

./restart.sh

This script automatically:

• Cleans up previous instances
• Activates the Python virtual environment
• Starts the web server at http://localhost:8000
• Initializes the game with Ollama integration

LangGraph Workflow: The game uses a sophisticated AI agent system where each agent has specialized tools and responsibilities. Agents communicate through a shared game state and make decisions based on their observations.

Web Interface: Built with vanilla JavaScript and modern CSS, featuring real-time data polling for live updates of AI agent decisions and game state changes.

Local LLM Integration: Utilizes Ollama for private, local language model processing, ensuring data privacy while maintaining powerful AI capabilities.

• Objective: Collect all treasures while avoiding or defeating enemies
• Movement: Ship can move up to 3 tiles per turn (North/South/East/West)
• Combat: Cannons have 2-tile range and cost 1 cannonball per shot
• Resources: Start with 25 cannonballs, gain 2 per treasure collected
• Hazards: Contact with enemies or monsters causes damage
• Victory: Collect all treasures to win the game

Contributions are welcome! Areas for enhancement:

• New AI agent strategies and behaviors
• Additional map types and game modes
• Enhanced web interface features
• Performance optimizations
• Mobile responsiveness improvements

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

• LangGraph: For the powerful AI agent framework
• Ollama: For local LLM capabilities
• Material Design: For the clean icon system
• Chrome DevTools: For the innovative workspace integration

Ahoy, matey! May the winds fill your sails and the AI guide your treasure hunt! ⚓🏴‍☠️💰