Submit, poll, read: what a long analysis looks like over HTTP
Nothing here schedules anything
Page 3 of 4
The most common misreading of this shape is that submitting is the same as scheduling. It is not. The platform has no scheduler. There is nothing to configure that says "run this analysis at 02:00", and there is no cron page waiting to be built.
Applications run when something calls them, and that something is yours: the job that already knows your source systems have landed, the workflow tool your team already operates, the person clicking a button. That is the right place for it. The only party that knows the nightly extract finished is the system that finished it, and a scheduler on our side would be guessing at a time and being wrong on every month-end.
So the pattern in practice is: your scheduler fires, calls POST /ops/<key>, keeps the job id, polls
until the job is done, and reads the output. Four lines of whatever you already have.