Visualizes the end-to-end process: a user prompt enters the model, is processed, and the model streams back a generated response token by token.