Adding AR term to ForecastingModel

Hi, I’m following time series forecasting tutorial using ForecastingModel

I’m trying to add an explicit auto-regressive term to my model, but since only zero_data is provided, how could I do that in an elegant way? I know I can do it by passing data in a workaround, i.e global variables or passing the lags as a covariates, etc. Following up on that, why was forecasting module designed to take only the shape of the data? I was trying to figure it out in the tutorials and documents but all I found was “Since this is a generative model, it shouldn’t look at the data”. I’m curious about that design principle.

Thanks in advance for your help and I appreciate any further suggestions and advice.