I want to evaluate forecasting accuracy of my model. so, I am using filter_step_ahead option. It works, however, I want to evaluate conditional forecasting accuracy with given some external variables.

In other words, assume that model is
y=f(y(-1),x)+e ← domestic economy
x=AR(1) ← external sector

I need conditional forecasts of y_t+1 for any t in a range 2:T-1, when model knows y_1:y_t and whole path of x_1:x_T.

I have a data for a range 1:T, for each t=2:T-k, I want my model to smooth data in a range 1:t and make conditional forecast of y for a range t+1:t+k, conditional to x_(t+1:t+k), assuming that innovations of x in forecast range are unknown in the period t.