Recipe API

The Recipe API is the YAML interface for full forecasting studies. Use it when the high-level Simple API is too small: multi-layer DAGs, explicit data policy, model comparisons, statistical tests, interpretation, and output provenance.

A recipe is a top-level mapping whose layer keys are:

0_meta, 1_data, 2_preprocessing, 3_feature_engineering, 4_forecasting_model, 5_evaluation, 6_statistical_tests, 7_interpretation, 8_output.

Diagnostic layers may also appear:

1_5_data_summary, 2_5_pre_post_preprocessing, 3_5_feature_diagnostics, 4_5_generator_diagnostics.

Start here:

• Recipe Gallery: runnable examples.

• Recipe Layer Contract: layer keys, DAG node shape, and one complete recipe.

• Data Layer: source, target, horizon, and FRED-SD geography choices.

• Data Policies: missingness, outliers, release lags, and same-period predictors.

• Recipe Defaults: package-level defaults from macroforecast.defaults.

• Runtime Support Matrix: what executes today.

• Understanding Recipe Output: artifact directory and manifest layout.

• FRED Datasets in Recipes: FRED-MD, FRED-QD, and FRED-SD reference status.