All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Report chart no longer skews on error scores (999 / Infinity) — they are replaced with
nullso Chart.js renders a clean gap - Error iterations now show an orange "Error" badge and auto-expand benchmark output, making failures immediately visible
- Score display shows "Error" instead of raw sentinel values (999.0 / Infinity)
- Iteration reasoning is now labeled "Hypothesis" to clarify it is the LLM's plan before the benchmark ran
- Error stdout styled in orange to distinguish from successful output
--goal-fileoption: pass a file path instead of an inline goal string, enabling rich, multi-line goal descriptions with as much context as needed. Fully backwards-compatible — the positionalGOALargument works exactly as before.
- Initial public release with autonomous optimization loop, snapshot strategies (
copy,git), budget cap, dry-run mode, and multi-model support (Claude, OpenAI, Ollama).