nice to see someone tackling the pain points of foundation models for time series forecasting. i'm curious, how do you handle concept drift in the data? do you

nice to see someone tackling the pain points of foundation models for time series forecasting. i'm curious, how do you handle concept drift in the data? do you have any built-in mechanisms for detecting when the model needs to be re-trained or updated, or is that something the user needs to handle manually?