Weekly episode on a schedule: the $1.00 default cap stops video

Why does a weekly scheduled episode stop short? A schedule's generation spend cap defaults to $1.00 when you do not set one, and an API caller can lower the cap for a run but never raise it. Any episode whose generation costs more than $1.00 will stop at the default.

The weekly slot is how most series are run: a new episode every Thursday. Scheduled is the Sume product for that cadence, and its spend ceiling is the part to configure first.

The cap rules

A schedule has instructions, a model, a cron expression and a spend cap, and it runs as an Agent in a fresh thread. You create and edit schedules in the Agents dashboard or by asking the Agent in chat; the Developer API can list, read and start runs but cannot create or edit a schedule.

Overlap and the slow week

Only one run of a schedule is active at a time. on_active_run is skip (the default) or reject: skip records a skipped run, reject answers 409 action_run_in_progress. A video episode that runs long into the next slot is the case this decides, so pick reject if a skipped week is worse than a retried one.

A schedule run is not a job. It does not appear in /v1/jobs; read it at /v1/action-runs/{run_id}. The names are historical: the namespace is /v1/actions and ids start with aut_.

Setup checklist

Set the cap to what one episode really costs plus margin. Decide on skip or reject. Bind an output schema so each run returns the episode record. Use a Format instead when your own backend, not the clock, starts the work.

Picking a number

A good cap is the per-episode cost plus a margin. Price one episode from the live rate card, then add room for a retry. Set it when you build the schedule, since the API cannot create one for you. If one week needs less, the caller can pass a lower number on that run, never a higher one.

• Set the cap to the real episode cost plus margin.
• Use reject if a skipped week is worse than a retry.
• Read the result at /v1/action-runs/{run_id}.

When to use a Format instead

Scheduled is the right fit when the clock starts the work. If your own backend decides when an episode renders, call a Format directly; it gives you idempotency keys, webhooks and a per-run spend cap. You can combine them: a schedule that drafts the outline and a Format that renders approved episodes on demand.