I’m new to temporal. We are considering modeling out a AI chat interaction with a temporal workflow
User: what is the capital of France
Ai: Paris
<—
User: who is their leader?
I’m struggling if this moment represented by the arrow where we return control to the user is better represented as a signal or a completed workflow.
signal thoughts
Chats can live forever. I can go back to a chat from 4 months ago and add a new message. Is there limits to the number of jobs that can stay waiting? Are there limits on how big timeouts can be?
complete workflow thoughts
This would be the version where each new user message in the chat would be a new workflow returning to the user when the workflow completes. No history problems since I think complete workflows are scalable. Id need to manage long lives state out side temporal but maybe I need to do that anyway because of the size limitations of messages.