feat(pipelines): compute_harmonization_diagnostics — long-format pre/post ComBat for viz

feat(frontend): trust caption — precision-at-confidence below decision card

feat(api): expose calibration bin in /predict/bbb response

feat(models): calibration metadata — precision-at-threshold bins on _neurobridge_calibration

feat(frontend): edge-case dropdown for BBB robustness demo

docs: Day-5 close-out — AGENTS §8 decision layer + trainer CLI

feat(frontend): interactive BBB tab — SMILES input + decision card + SHAP bars

feat(api): POST /predict/bbb with prediction, uncertainty, SHAP top-k

feat(models): SHAP top-k explainer for BBB predictions

feat(models): BBB classifier with predict_with_proba uncertainty

feat(frontend): Streamlit dashboard with 3 modality tabs

feat(api): POST /pipeline/{bbb,eeg,mri} dispatch routes

feat(api): scaffold FastAPI app + /health + shared Pydantic schemas

feat(mri): log run params, metrics, and parquet artifact to MLflow

feat(eeg): log run params, metrics, and parquet artifact to MLflow

feat(bbb): log run params, metrics, and parquet artifact to MLflow

feat(core): add MLflow tracking helper with disable env-flag

refactor(bbb): align pin_threads() comment with EEG/MRI sibling files

refactor(mri): use core.determinism + core.storage helpers

refactor(eeg): use core.determinism + core.storage helpers

refactor(bbb): use core.determinism + core.storage helpers

feat(core): extract write_parquet() helper for §6 storage contract

feat(core): extract pin_threads() helper for determinism

docs: fix README cross-references; type _StatFn as a real Callable alias

fix(mri): handle all-constant features; tighten variance threshold; reorder log

feat(mri): add run_pipeline orchestrator + CLI (NIfTI dir → ComBat Parquet)

fix(mri): pin ComBat transitive deps; clarify np.round(14) intent; guard inputs

feat(mri): add harmonize_combat wrapper around neuroHarmonize.harmonizationLearn

refactor(mri): bind ROI_STATS to callables; guard volume/mask shape mismatch

feat(mri): add extract_features_from_volume (8 ROI octants × 6 stats)

fix(mri): warn on all-False mask; document 6-connectivity erosion caveat

feat(mri): add mask_brain (intensity threshold + morphological opening)

feat(mri): add is_valid_volume guard for NaN/inf/shape/dtype on 3-D arrays

refactor: pin single-threaded determinism env; close Day-2 doc/typo gaps

docs(eeg): correct float64 dtype claim; tighten test match; clarify CLI default

feat(eeg): add run_pipeline orchestrator + CLI (FIF/EDF → Parquet)

fix(eeg): NaN-clean features for flat channels; guard zero-size epochs; assert WARNING

fix(eeg): validate epochs after filter to guarantee no NaN in feature table

feat(eeg): flatten 3D epochs into deterministic 2D feat_<ch>_<band|stat> table

refactor(eeg): bind STATS labels to callables; document moment conventions

feat(eeg): add compute_features_from_epoch (PSD bands + 5 statistics)

refactor(eeg): tighten ICA docstring + log accuracy; extract EOG threshold constant

feat(eeg): add remove_artifacts_with_ica with EOG correlation rejection

fix(eeg): guard bandpass_filter against inverted l_freq/h_freq; document picks=all

feat(eeg): add non-mutating bandpass_filter (default 1-40 Hz)

fix(eeg): tighten is_valid_epoch signature; reject non-numeric dtype + -inf

feat(eeg): add is_valid_epoch guard for NaN/inf/shape/dtype

docs(bbb): fix stale CSV mentions in run_pipeline + module docstrings

feat(bbb): switch processed output to Parquet for dtype preservation

refactor(bbb): pin LF lineterminator; move DEFAULTs to top; reject dir output