darwinian-evolver

Automate prompt engineering by running an LLM-driven evolutionary search loop that mutates and scores prompt candidates against a defined fitness function. This iteratively refines prompts over multiple generations to maximize success rates against your test suite.

You can optimize SQL queries and regex patterns automatically by defining custom organisms and evaluators in Python. The evolutionary loop mutates these artifacts and scores them against measurable fitness functions to find optimal candidates.

Evolutionary optimization for code snippets uses an LLM-driven loop to mutate code and evaluate it against a fitness function. It applies heuristic-based iterative refinement to discover improved code variants across multiple search generations.

Yes, you need OpenAI and Jinja2 dependencies installed to run the driver scripts. These libraries provide the LLM integration for mutation operations and the templating engine required to define custom problem organisms and evaluators.

Define a custom fitness function by using the provided templates to specify your own organisms, evaluators, and mutators. This allows the evolutionary search loop to score domain-specific artifacts like prompts or queries against your unique test criteria.

Avoid using an evolutionary loop for prompt optimization when your task lacks a measurable fitness function or sufficient test cases. The iterative mutation and scoring process requires quantifiable evaluation criteria to effectively guide the search toward optimal candidates.