Capabilities
Features
Everything one trained model does — grouped into four areas, each with what it means, when to reach for it, and the code to use it.
01
Forecasting & uncertainty
The questions one trained model answers, and the calibrated intervals behind them.
Grid forecast Forecast at any moment When will it cross the line? Schedule a whole fleet Conformal intervals Calibrated event-time windows
Explore →
02
Inputs & features
Go beyond a single series: sensors, known drivers, categories, operating regimes.
Multivariate sensor fusion Exogenous variables Categorical features Regime normalization
Explore →
03
Evaluation & monitoring
Score the way production will, then watch for drift after you ship.
Rolling-origin backtesting A full metric suite Drift & coverage monitoring
Explore →
04
Production & architectures
Persist, export, serve and reproduce — plus the model choices under the hood.
Persistence TorchScript export FastAPI serving Reproducible runs Experiment logging sklearn-compatible Typed Command line Neural ODE Transformer TCN
Explore →
New to APDTFlow?
Start with the Quickstart for the 60-second tour, see the capabilities working in the Demos, or read the full usage guide and capabilities reference on GitHub.