Simulated hub history and a next-day meal forecast. The plan cuts food waste and still keeps a service buffer.
A catering hub that boards meals for 300 or more departures a day needs two views. One is history: trays loaded, trays eaten, and whether the cart left with enough time before departure. The other is tomorrow: how many trays to produce for each flight number and meal type. CaterFlow is a batch pipeline, not an HTTP service. generate_data.py writes the history. forecast_engine.py writes the plan. Power BI reads both.
QR1001-20261005Review
A seed-42 run through 2026-10-25 produces 6,720 flights, about 14.5% historical waste, and a Monday plan of 111,683 trays against 125,718 recently loaded. production_delta is the cut or the add versus recent loading. The waste reduction figure is a planning estimate, not a measured result.
Seat maps, the partner-airline mix, and the route list are planning assumptions, not an official catering feed. The full write-up covers the schema, both CLIs, and the forecast columns.