The AI suggestion service will be returning a suggestion through the API via a websocket meaning real time suggestion, how to best display that text to the user? One word at a time, all at once, a block that shows it's loading or processing during the latency?