Add a fake model that always returns the same canned response regardless of what the prompt was.

Add a fake model that always returns the same canned response regardless of what the prompt was. #118