Skip to content

Releases: RNBBarrett/AutoEvolve

v0.1.0 -- Initial Release

05 Apr 18:47

Choose a tag to compare

autoevolve v0.1.0

First public release of autoevolve -- an evolutionary optimization framework powered by LLM-guided mutation.

Highlights

  • Evolve Python code, prompts, or any text artifact
  • Three built-in strategies: beam, beam_archive, swarm
  • Four mutation modes: rewrite, patch, nl_feedback, crossover
  • Five LLM providers: mock, anthropic, openai, ollama, external_command
  • OPRO-style natural language feedback for history-aware mutation
  • Subprocess-based evaluator sandboxing
  • Archive with top-k, recent-k, and diverse-k selection
  • Run resume support
  • Three bundled examples with step-by-step documentation
  • Comprehensive test suite

Getting Started

pip install -e .
python -m autoevolve run --task examples/01_text_compression_poem/task.yaml

See the README for full documentation.