Skip to content

Customization

Learn how to extend EvoToolkit to solve your specific optimization problems.

Customization Options

Custom Tasks

β†’ Start Tutorial

Create your own optimization tasks for domain-specific problems.

You'll Learn: - Extending the Task base class - Implementing custom evaluation logic - Defining solution spaces - Integrating with evolutionary algorithms

Prerequisites: Basic EvoToolkit knowledge (Scientific Regression tutorial)

Customizing Evolution Methods

β†’ Start Tutorial

Learn how to customize evolutionary behavior by modifying prompts or developing new algorithms.

You'll Learn: - Understanding the Interface architecture - Customizing LLM prompts to improve results - Designing task-specific prompts - Developing brand new evolution algorithms (advanced) - Implementing custom strategies like temperature annealing

Prerequisites: Basic EvoToolkit knowledge (Scientific Regression tutorial)

When to Customize

Create a Custom Task When:

  • Your problem domain isn't covered by built-in tasks
  • You need specific evaluation metrics
  • You want to optimize domain-specific code or structures
  • You need custom constraints or validation

Customize Evolution Methods When:

  • Default prompts don't work well for your task
  • You want to incorporate domain knowledge into the evolution
  • You need special mutation or crossover strategies
  • You want to experiment with novel evolutionary approaches

Quick Start Examples

Custom Task Example

from evotoolkit.core import BaseTask, Solution

class MyCustomTask(BaseTask):
    def evaluate(self, solution: Solution) -> float:
        # Your evaluation logic here
        return fitness_score

Custom Interface Example

from evotoolkit.task.python_task import EvoEngineerPythonInterface

class MyCustomInterface(EvoEngineerPythonInterface):
    def get_prompt_components(self):
        # Customize prompts for your task
        return custom_prompts

Best Practices

  1. Start Simple: Begin with modifying existing tasks or prompts before creating entirely new ones
  2. Test Thoroughly: Validate your custom evaluation logic with known solutions
  3. Document Well: Clearly document your task requirements and constraints
  4. Share Your Work: Consider contributing your custom tasks back to the community

Next Steps

After mastering customization: - Explore Advanced Usage for low-level API control - Check the API Reference for detailed documentation - Share your custom tasks in GitHub Discussions