Running a task is roughly:
• fetch current task details from server
• Get all the machines to fetch the files from the server and verify all get the same checksum. Do this in batches
• If all machines have the same task cached (so we know no one updated the task mid run causing weirdness) execute it and wait a while
• If it’s not done after a while give details so progress can be checked on later.
So task REST api, inspecting result, choria request, choria request potentially more choria requests and custom UX wrapping it all together.
That's what an “application” is - more than a request it's all the extra logic and UX flow and orchestration