
Prompt Optimization with GEPA
The open-source GEPA prompt optimizer. Feed GEPAzilla a dataset and it will chew through prompts, score them with custom metrics, and spit out the strongest contender—feng shui for your system prompt.
Bring your dataset
Load your own rows or start with data/sample-dataset.json
to see GEPAzilla in action.
Composable scoring
Mix deterministic checks with LLM judges. Weight and duplicate scorers to match your eval stack.
Local telemetry
Every span stays on your machine—tokens, routing, costs, and errors are yours to inspect.