LLM-powered mutations
Language models generate intelligent variations instead of random perturbations, accelerating convergence.
LLM Evolution
Evolutionary algorithms powered by LLMs
Python 3.12+ · Protocols · Generics2025Alejandro Fernández Camello
A Python library that enables Large Language Models to act as intelligent mutators, crossover operators, and evaluators within evolutionary algorithms.
Evolution has no goal, only a process. We provide the goal; the LLM provides the intelligence.
Evolution in action
Watch a population evolve generation by generation toward the optimal solution.
Generation: 0/200Best fitness: 0%Avg fitness: 0%
Target
Features
Modular components to evolve any type of solution.
Protocol-based design
Clean interfaces built on Python Protocols let you swap strategies without tight coupling.
Generic framework
Evolve anything: from simple integers to CUDA kernels or algorithmic trading strategies.
Modular components
Separate strategies for initialisation, evaluation, selection, crossover, and mutation.
Domain agnostic
Apply it to CUDA, RISC-V assembly, algorithmic trading, or any optimisation problem.
Open source
MIT licensed. Contribute, extend, and adapt the library to your needs.
Explore the code
Open-source library ready to use in your LLM-powered optimisation projects.