Godot LLM Framework

Table of Contents

• Introduction
• Features
• Installation
• Usage
  • Basic Setup
  • Generating Responses
  • Using Tools
  • Creating Custom Tools
• API Reference
• Configuration
• Examples
• Contributing
• License

Introduction

The Godot LLM Framework is a powerful addon for the Godot game engine that enables seamless integration of Large Language Models (LLMs) into your game development workflow. This framework provides a unified interface for working with different LLM providers, making it easy to incorporate advanced natural language processing, dialogue generation, and dynamic content creation directly within your Godot projects.

Features

• Support for multiple LLM providers (currently Anthropic's Claude, with OpenAI support in progress)
• Easy-to-use API for generating responses from LLMs
• Tool system for extending LLM capabilities with custom functions
• Configurable settings for fine-tuning LLM behavior
• Message history management for contextual conversations
• Asynchronous operations for smooth integration with Godot's event loop

Installation

Find the addon using the Godot Asset Library or follow the instructions below for manual installs

1. Download the godot_llm_framework folder from this repository.
2. Place the folder in your Godot project's addons/ directory.
3. Enable the addon in your project settings:
   • Go to "Project" > "Project Settings" > "Plugins"
   • Find "Godot LLM Framework" and check the "Enable" box

Usage

Basic Setup Scene Tree

1. Add an LLM node to your scene.
2. Configure the LLM node in the Inspector:
   • Set the Provider (e.g., ANTHROPIC)
   • Enter your API key
   • Adjust other settings as needed (model, temperature, etc.)

Basic Setup GDScript

var llm_config = LLMConfig.from_dict({
    "provider": LLMProviderAPI.Provider.ANTHROPIC,
    "api_key": "Your API key for the chosen provider",
    "model": "claude-3-haiku-20240307",
    "temperature": "0.7",
    "max_message_history": 20
})

var llm = LLM.create(llm_config)

Generating Responses

@onready var llm = $LLM

func _ready():
    var response = await llm.generate_response("Tell me a joke about game development.")
    print(response)

Using Tools

1. Register tools with your LLM instance:

llm.add_tool(WeatherTool.new())
llm.add_tool(UserNameTool.new())

2. Generate responses that may use tools:

var response = await llm.generate_response("What's the weather like in California, and can you call me by my name?")
print(response)

Creating Custom Tools

1. Create a new script that extends BaseTool:

extends BaseTool

class_name MyCustomTool

func _init():
    super._init(
        "my_custom_tool",
        "Description of what my tool does",
        {
            "type": "object",
            "properties": {
                "input_param": {
                    "type": "string",
                    "description": "Description of the input parameter"
                }
            },
            "required": ["input_param"]
        }
    )

func execute(input: Dictionary) -> String:
    # Implement your tool's functionality here
    return "Result of my custom tool"

2. Register your custom tool with the LLM:

llm.add_tool(MyCustomTool.new())

API Reference

For detailed API documentation, please refer to the inline comments in the source code, particularly in the following files:

• llm.gd
• llm_provider_api.gd
• llm_config.gd
• base_tool.gd

Configuration

The LLM's behavior can be configured through the LLMConfig resource. Key settings include:

• provider: The LLM provider to use (e.g., ANTHROPIC)
• api_key: Your API key for the chosen provider
• model: The specific model to use (e.g., "claude-3-haiku-20240307")
• temperature: Controls the randomness of the output (0.0 to 1.0)
• max_message_history: The maximum number of messages to keep in the conversation history

Examples

Check out the example.gd and example.tscn files in the addon folder for a working example of how to use the Godot LLM Framework.

Contributing

Contributions to the Godot LLM Framework are welcome! Please follow these steps:

1. Fork the repository
2. Create a new branch for your feature or bug fix
3. Commit your changes with clear, descriptive commit messages
4. Push your branch and submit a pull request

Please ensure your code follows the existing style and includes appropriate documentation.

License

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