Skip to content

Advanced Usage

Master the low-level API for maximum control and customization.

Overview

These advanced tutorials cover:

  • Low-Level API - Direct algorithm control and configuration
  • Algorithm Configuration - Fine-tune evolution parameters
  • Algorithm Internals - Access and analyze internal state
  • Debugging & Profiling - Troubleshoot and optimize performance

Prerequisites

Tutorials

Low-Level API

→ Start Tutorial

Learn the difference between high-level and low-level APIs, and when to use each.

You'll Learn: - High-level vs low-level API comparison - Direct algorithm instantiation - Accessing internal state - Custom workflow control

Time: 15 minutes

Algorithm Configuration

→ Start Tutorial

Master detailed configuration options for each evolutionary algorithm.

You'll Learn: - EvoEngineer configuration parameters - FunSearch island model setup - EoH operator control - Parallel execution tuning

Time: 20 minutes

Algorithm Internals

→ Start Tutorial

Access and analyze the internal state of evolutionary algorithms.

You'll Learn: - Inspect evolution history - Access solution populations - Plot evolution progress - Extract metrics and statistics

Time: 15 minutes

Debugging & Profiling

→ Start Tutorial

Debug issues and optimize performance of your evolutionary workflows.

You'll Learn: - Enable verbose logging - Save intermediate solutions - Inspect LLM prompts/responses - Time and memory profiling - Implement custom algorithms

Time: 25 minutes

When to Use Advanced Features

Use Low-Level API When:

  • You need fine-grained control over the evolution process
  • Default configurations don't meet your requirements
  • You want to implement custom stopping criteria
  • You need access to intermediate results

Use Custom Configuration When:

  • Default parameters don't work well for your task
  • You want to optimize for speed or quality
  • You need to tune parallel execution
  • You're experimenting with algorithm variants

Use Debugging Tools When:

  • Evolution doesn't converge as expected
  • You want to understand algorithm behavior
  • You need to optimize resource usage
  • You're developing custom algorithms

Quick Reference

Basic Low-Level Pattern

from evotoolkit.evo_method.evoengineer import EvoEngineer, EvoEngineerConfig

config = EvoEngineerConfig(
    interface=interface,
    output_path='./results',
    running_llm=llm_api,
    max_generations=10,
    pop_size=8,
    verbose=True
)

algorithm = EvoEngineer(config)
algorithm.run()

# Access results
best = algorithm._get_best_sol(algorithm.run_state_dict.sol_history)

Next Steps

After mastering advanced usage: