v2.1.0

[2.1.0] - 2026-06-10

• [FEAT] New model inspection API: model.summary() and model.info() expose model metadata, training and dataset statistics, and model-selection criteria (AIC, BIC, ICL). The returned objects auto-print when their result is discarded and stay silent when assigned to a variable (see #485)
• [FEAT] Mixture model completed: remaining clustering functionality and ICL-based evaluation (see #489)
• [DOC] New developer documentation: architecture (DAG) diagrams, model animations, contribution workflow, and a release guide (see #476)
• [DOC] Documentation example gallery modernized and trimmed (quickstart, Parkinson, joint, mixture, simulate) (see #485, #494, #496)
• [DOC] Fixed Sphinx / ReadTheDocs build and numerous documentation warnings (see #483, #490)
• [FIX] Broad bug fixes across models (base, joint, linear, logistic, lme, mixture) and the fit output manager (see #494)
• [FIX] Numerical and warning fixes: ICL no longer produces NaN (and corrupts JSON export) when a cluster probability reaches exactly 0 (convention p·log(p) → 0); use_jacobian defaults to None ("auto") to avoid a spurious warning on fit/personalize; and scalar-noise warnings and issue #501 are resolved
• [CHANGE] simulate now groups warnings and defaults to verbose_warnings=False, printing a summary count with a hint to set verbose_warnings=True for details

What's Changed

• Merge branch v2.1 to master, Release by @MendezSebastianP in #502

Full Changelog: 2.0.2...v2.1.0