                 Changes in version 0.2.0 (2026-09-05)                  

  - Added calibration plots (binned or smoothed) and Brier score to
    tr_validate(), closing a TRIPOD+AI reporting gap. New
    calibration_method argument ("binned" or "smooth").
  - Calibration plot now included in tr_tripod_report() output.
  - Added robustness handling for models with limited probability
    variation (e.g. tree-based models with many identical predictions).
  - Added survival analysis support: tr_fit_survival() (Cox Proportional
    Hazards and Random Survival Forest via the censored/aorsf engines)
    and tr_validate_survival() (concordance index / C-index).
  - Added tr_launch_app(): an interactive Shiny application for the full
    triageR workflow (upload data, fit and compare models, validate,
    explain, review, and download a TRIPOD+AI report) without writing R
    code.
  - Fixed test suite reliability issues related to lazy-loaded survival
    datasets.

                 Changes in version 0.1.1 (2026-08-02)                  

*PimaIndianDataset in the mlbench package was removed in the latest
update, thus the vignette now uses the inbuilt MASS which still has the
Pima Indian Diabetes Dataset.

                 Changes in version 0.1.0 (2026-07-29)                  

  - Initial development version.
  - Added data layer: tr_load_clinical(), tr_check_missing(),
    tr_impute().
  - Added model layer: tr_fit(), tr_validate(), tr_explain().
  - Added agent layer: tr_recommend_method(), tr_sensitivity(),
    tr_agent_review().
  - Added reporting layer: tr_tripod_report().
