Can a tf-agents environment be defined with an unobservable exogenous state?

Question

I apologize in advance for the question in the title not being very clear. I\'m trying to train a reinforcement learning policy using tf-agents in which there exists some un