module Docs::A_GETTING_STARTED::EVOLUTION_GUIDE

Overview

Feedback-Driven Evolutionary Optimization

Crowbar enables programs to report candidate execution feedback (success/fitness) to evolve inputs toward maximizing coverage, depth, or custom test goals.

Executive Summary & Key Topics

Topic	Method / Anchor	Description
Genetic Algorithms and Diversity Controls	`.topic_01_evolution_mechanics`	Population selection, crossover, and preventing local locking.

Related Guides & Source References

src/crowbar/evolution/manager.cr
src/crowbar/evolution/corpus.cr

Defined in:

docs/a_getting_started/evolution_guide.cr

Class Method Summary

.topic_01_evolution_mechanics : Nil
Genetic Algorithms and Diversity Controls: Population selection, crossover, and preventing local locking.

Class Method Detail

def self.topic_01_evolution_mechanics : Nil #

Genetic Algorithms and Diversity Controls: Population selection, crossover, and preventing local locking.

Parameter	Default	Purpose
`enabled`	`true`	Toggle evolutionary mode on/off
`population_size`	`64`	Maximum candidates preserved in memory
`exploration_rate`	`0.15`	Percentage of forced random baseline inputs
`crossover_rate`	`0.25`	Percentage of hybrid offspring recombination
`stagnation_limit`	`100`	Stagnant cycles before novelty culling

Working Examples

fuzzer = Crowbar.define do
  evolution do
    enabled true
    population_size 32
    exploration_rate 0.15
  end
end

1000.times do
  candidate = fuzzer.fuzz(baseline)
  # Evaluate candidate in target program
  fitness_score = evaluate(candidate)
  fuzzer.report(candidate, fitness: fitness_score)
end

crowbar