darwinian-evolver

Automated prompt optimization uses an LLM-driven evolutionary loop to mutate and score text artifacts against a defined fitness function. This automates the search for high-performing variants, replacing manual tuning with iterative refinement against measurable criteria.

Optimizing underperforming prompts requires defining organisms, evaluators, and mutators for your specific domain. You evolve the prompt template by running the evolution loop for multiple iterations until it achieves a high success rate on your evaluation criteria.

Yes, executing LLM-driven evolutionary search requires Python 3.11, git, uv, and an LLM provider API key. These dependencies are necessary to run the mutation and evaluation cycles for automated refinement.

Yes, evolutionary search optimizes text-based artifacts like regex patterns and code snippets. It applies mutators and evaluators to these artifacts, scoring candidates against a fitness function to automate refinement for research and development tasks.

An LLM mutates and scores candidates by applying a custom mutator to alter text, then using an evaluator to measure performance against your fitness function. Candidates achieving high scores survive, driving iterative optimization across generations.

LLM-driven mutation for SQL generation requires defining a strict fitness function to evaluate query accuracy. Without measurable evaluation criteria and sufficient API calls for multiple iterations, the evolutionary loop may fail to converge on optimal logic.