Hey, what's the industry standard way to continuously stream text from an reasoning AI that can take longer than 10minutes? Currently using @convex-dev/agent with an action.
My current workflow has multiple steps, but some steps has deep research tasks that cannot be split up anymore. The problem is that a step can take more than 10 minutes which means that the convex action finishes before we get the final result and we never finish streaming the text.
Currently steps that finish within 10 minutes work, and we can see live text streaming in the FE. But it will stop once a step goes over 10 minutes
The only way I can think of getting round it is to do it off convex and keep updating the table.