Is there a possiblity that we can get a callback on what the model is thinking before starting the output? the copilot cli shows that in gray.