[embedding-shape]

Not all models are trained with long one-shot task following by themselves, seems many of them prefer closer interactions with the user. You could always add another layer/abstraction above/below to work around it.

[fastball]

Can't this just be a Ralph Wiggum loop (i.e. while True)

[embedding-shape]

Sure, but I think just about everyone wants the agent to eventually say "done" in one way or another.