darwinian-evolver

You can optimize prompts and code by defining an Organism, Evaluator, and Mutator to run an LLM-driven evolutionary loop. This loop applies a fitness function to score candidates and proposes mutations to iteratively improve performance.

Evolutionary search for regex and SQL queries is a process where an LLM mutates candidate expressions to improve their performance. It uses trainable and holdout sets to score fitness and guard against overfitting during iterative improvements.

Yes, you can use OpenRouter and OpenAI providers to evolve artifacts. The framework includes a built-in parrot driver and supports custom drivers to integrate multiple LLM providers for the evolutionary search loop.

Yes, Python 3.11 or newer is required to run the LLM-driven evolution workflow. You also need an LLM provider like OpenRouter or OpenAI and the framework repository included in the skill.

To prevent overfitting when evolving prompts and regexes, the framework leverages separate trainable and holdout sets. This safe workflow ensures candidates generalize well and provides snapshots for later analysis.

The best way to start evolving small code snippets is by running the parrot_openrouter example. This quick start example demonstrates how to perform an iterative evolution of an artifact end-to-end.