Spread Graph API calls evenly: pace a nightly Reel batch

Meta's Graph API guidance for staying under its limits includes spreading queries evenly to avoid traffic spikes, and its call_count is measured over a rolling hour. For a nightly Reel batch, divide the hour by the number of publish calls and sleep between them, instead of firing the whole batch the moment renders finish.

Pace the render side separately: Sume reports generation_limits on submit responses, and its docs describe wave_size_hint as a submission-wave hint, not a concurrency limit.

Two budgets, two pacers

Keep the render pacer and the publish pacer independent.

How do I space the calls?

Compute the gap once, then sleep. The sample spaces a list of items across a window.

import time


def spaced(items: list, window_seconds: int):
    gap = window_seconds / max(1, len(items))
    for index, item in enumerate(items):
        if index:
            time.sleep(gap)
        yield item


if __name__ == "__main__":
    for job_id in spaced(["job_a", "job_b", "job_c"], window_seconds=3):
        print("publish", job_id)

What about the render wave?

Do not size in-flight Sume work from wave_size_hint; use the effective concurrency_limit minus active and queued jobs, capped by queue_capacity_remaining (Generation admission). The post on pacing bulk Sume submits has the full loop.