Customization
Customization Overview
Understand how CUGA can be customized
You can customize CUGA's behavior and features through the use of configuration files and environment variables.
Configuration Structure
By default, CUGA uses configuration files which are stored in the src/cuga/configuration/ folder, however you can
specify a custom configuration directory via this environment variable:
# Set custom configuration directory
export CUGA_CONFIGURATIONS_DIR="/path/to/your/configurations"# Set custom configuration directory
setx CUGA_CONFIGURATIONS_DIR "C:\path\to\your\configurations"This allows you to organize multiple configurations in separate folders.
Each configuration folder, contains TOML and MD files that allow to toggle features and customize behavior.
src/cuga/configurations/
├── models/
│ ├── settings.litellm.toml # Default ETE configuration
│ ├── settings.openai.toml # OpenAI configuration
│ ├── settings.watsonx.toml # WatsonX configuration
│ └── settings.azure.toml # Azure configuration
│ └── settings.groq.toml # Groq configuration
├── modes/
│ ├── fast.toml # Fast mode settings
│ ├── balanced.toml # Balanced mode settings
│ └── accurate.toml # Accurate mode settings
└── instructions/
└── default/
├── api_planner.md # Custom API planner instructions
├── task_decomposition.md # Custom task decomposition instructions
└── code_agent.md # Custom code agent instructionsCustomization topics
The following pages include additional information on specific customzation topics:
- Configure your LLM providers of choice
- Configure environment variables and system settings
- Customize reasoning modes and advanced features for optimal performance
- Configure CUGA's task execution modes
- Specify Special Instructions for different sub agents
- Onboard tools or integrate with existing tool runtime environments