"""Model training pipelines for Learning Outcome OS AI Services V2.

This package contains offline training scripts for all 7 AI models:
- LO Tagger (TF-IDF + CalibratedClassifierCV/LinearSVC)
- Bloom Classifier (TF-IDF + LogisticRegression)
- Difficulty Model (RandomForestRegressor)
- Mastery Model (RandomForestClassifier)
- Risk Model (RandomForestClassifier with balanced class weights)
- Answer Scorer (TF-IDF + Ridge regression)
- Recommender (GradientBoostingClassifier)

Each pipeline inherits from BaseTrainer, which enforces artifact layout,
split discipline, deterministic seeding, and metrics schema.

This package imports from app/data/ and app/core/ but is NEVER imported by app/.
Training is executed via `python -m training.train_all` or individual trainer scripts.
"""