v0.1.1

Highlights

Candidate Tree Visualization — GEPA now generates an interactive HTML lineage tree of all candidates explored during optimization. Nodes are color-coded by role (best, Pareto front, seed) with hover previews and click-to-pin tooltips for reading full prompt text. The tree is automatically logged to WandB and MLflow at each step — no configuration needed. (#256)

gskill: Automated Skill Learning for Coding Agents — A new gepa.gskill module that uses optimize_anything to automatically discover and learn repository-specific skills for coding agents, powered by SWE-smith and Docker. (#213)

Top-K Pareto Candidate Selector — New candidate_selection_strategy="top_k_pareto" limits parent selection to the top-K candidates by aggregate score (default K=5), focusing mutation effort on the most promising programs. (#246)

Improvements

• Parallel evaluation on by default — EngineConfig now defaults to parallel=True with max_workers=os.cpu_count(). No configuration changes needed. (#240)
• Atomic state saves — Checkpoints now use temp-file + os.replace(), preventing corruption on interrupted runs. (#242)
• JSON run artifacts — run_log.json and candidates.json are now written to the run directory after each state save, enabling post-hoc analysis without extra code. (#245)
• Cleaner experiment tracking — WandB/MLflow logging overhauled: engine config logged at start, best/seed prompts in summary, scalar metrics begin at step 0. (#243)
• Earlier callback firing — on_optimization_start now fires before seed valset evaluation, not after. (#244)

Bug Fixes

• cloudpickle is now a proper dependency in the full extras group — use_cloudpickle=True no longer raises ModuleNotFoundError. (#242)

Examples & Tutorials

• Four new optimize_anything example suites: AIME math, ARC-AGI, circle packing, and blackbox optimization. (#234)
• New tutorial on speeding up a Sudoku solver with GEPA. (#224)
• Pydantic AI tutorial added to showcase. (#211)

What's Changed

• Update blog post links with new URLs by @LakshyAAAgrawal in #220
• fix image social preview by @LakshyAAAgrawal in #221
• Update blog by @LakshyAAAgrawal in #223
• Add tutorial on speeding up Sudoku solver with GEPA by @vacmar01 in #224
• Add gskill: automated repository-specific skill learning for coding agents by @rohitsandadi in #213
• Add Sutro link to the documentation index by @LakshyAAAgrawal in #226
• docs: add Pydantic AI tutorial to showcase by @diego-lima in #211
• Add Google Scholar tags and Redirection for links by @LakshyAAAgrawal in #237
• add optimize-anything example suites by @lukeleeai in #234
• docs: add gskill guide by @rohitsandadi in #229
• Enable parallel evaluation by default in EngineConfig by @LakshyAAAgrawal in #240
• fix: add cloudpickle dep + atomic state saves by @LakshyAAAgrawal in #242
• docs: add BAML prompt optimization link and Bedrock FAQ by @LakshyAAAgrawal in #250
• feat: save run log and candidates as JSON to run_dir by @LakshyAAAgrawal in #245
• feat: add Top-K Pareto candidate selector by @LakshyAAAgrawal in #246
• fix: reduce WandB/MLFlow logging noise by @LakshyAAAgrawal in #243
• fix: fire on_optimization_start callback before seed valset eval by @LakshyAAAgrawal in #244
• docs: add community research papers citing GEPA by @LakshyAAAgrawal in #257
• feat: candidate tree visualization with interactive tooltips by @LakshyAAAgrawal in #256

New Contributors

• @rohitsandadi made their first contribution in #213
• @diego-lima made their first contribution in #211
• @lukeleeai made their first contribution in #234

Full Changelog: v0.1.0...v0.1.1