From warehouse schema to paper figures

The capstone notebook calls five numbered Python modules for schema discovery, feature construction, grouped modeling, descriptive query analysis, and reporting. DuckDB streams gated Parquet files; content-level outputs remain under ignored outputs/data/. Public aggregate records and the four figure exports are versioned with the site.

python -m venv .venv
pip install -r requirements.txt
set HF_TOKEN=<read-token>
python work/notebooks/01_signal_features_audited.py --cutoff 2026-04-30
python work/notebooks/01_signal_features_audited.py --cutoff 2026-05-31
python work/notebooks/02_modeling.py
python work/notebooks/03_query_associations.py
python work/notebooks/04_analysis_outputs.py

The grouped split and model sampling use seed 42. Download the capstone notebook. View the repository.