v1.1.0

1.1.0 (2026-10-07)

Features

• compile configured metric objectives (e1e3c4c)
• compile recipes into execution trees (d6a9af7)
• compose independent pipeline recipes (d43a10b)
• describe and compare recipes (c29f854)
• describe datasets without training (a4f2602)
• distinguish configured Bayesian search spaces (784f0af)
• edit recipes and component families (0443285)
• execute configured explanations on demand (4ef2684)
• expose the default pipeline as a recipe (885c7ec)
• honor explicit random seed domains (d2f03e4)
• introduce fixed values and numeric domains (73e994e)
• optimize declared training parameters (3eb4628)
• preserve metric aliases and report fold coverage (fb84207)
• reconstruct recipes as Python code (4a2ec1d)
• report resolved study snapshots (f82c744)
• restrict genetic swaps to declared alternatives (5e04552)
• track parameter policies in caches (cf7157b)
• train from attached pipeline recipes (141dda5)

Bug Fixes

• align binary curves with fitted classes (d1156f0)
• align permutation scoring with the main metric (1d94a57)
• compare final candidates on consistent evaluations (725422a)
• encode categorical features in minimal pipelines (c019e2e)
• handle open numeric bounds (97d09b4)
• honor configured MICE parameters (d6974a8)
• honor study configuration in baseline (139b24d)
• improve CatBoost hyperparameter search spaces (29f33b4)
• keep Sphinx caches outside published documentation (c81c1c3)
• make default feature selection optional (23f3583)
• preserve evaluations across sliding search stages (e4e4433)
• preserve inputs during inference (ad5062d)
• reduce survival forest memory usage (53c93f2)
• render box plots with the axes API (57d2e8b)
• respect automatic subsampling settings (1e9cc4e)
• support decision scores for binary AUROC (244304f)
• use supported MICE imputation fallback (757e296)

Performance Improvements

• skip unnecessary SVC probability calibration (321b962)
• start search with a lightweight baseline (ae75e94)
• use three folds for internal cross-validation (f534b19)

Documentation

• add basic recipe examples (227b4ca)
• add reusable strategies and study workflows (ed303a6)
• clarify stage and global search budgets (3cfde1e)
• connect guides and preserve previous URLs (c0cc94b)
• document recipe construction and editing (02fd394)
• document study analyses (7d558c5)
• introduce interactive pipeline discovery (5e38f2f)
• publish the Python example catalogue (c1ce297)
• update the README and existing user guides (bdc9521)