Skip to main content

​
Scheduled automations

Automations run work on a timetable: weekly digests, reminders, nightly syncs. You describe the outcome (who should get what, and when). GenieForge maps that to the schedule and action. The definition lives on the app; each project can enable it, override schedule or timezone, and run it manually. Find definitions under App Blueprint → Automations. With a project selected you also get project Automations for enablement and overrides.

​
How to verify

  1. Find the definition under App Blueprint → Automations.
  2. In the project, confirm Enabled, Timezone, and any Schedule Override.
  3. Use Run Now, then check Run History and project Executions.
  4. Confirm the wrong project does not receive another project’s digest.

​
App vs project

Use one project per customer or staging vs live-style boundary so digests and syncs do not cross data.

​
Timezones and schedules

On the project automation page, set Timezone (IANA presets such as America/New_York, or leave the server default). Use Schedule Override when this project should not use the app-level default schedule. Leave the override blank to keep the app default.

​
Running and history

  • Run Now on the project automation page triggers an immediate run (handy before launch).
  • Enable / Disable for Project controls whether the schedule fires there.
  • Run History (and app-level Recent Runs) show status such as Completed, Running, Failed, Pending, or Pending retry.
  • Open project Executions for broader task activity in that project.
Failed follow-up work should be safe to retry. Fix the underlying problem, then Run Now or wait for the next schedule. A failed digest should not invent that the source action never happened.

​
Credits on a run

Some automations run a fixed action with no AI turn (usually ~0 model credits). Others start a Chat-mode agent turn for digests or judgment work and cost model credits. GenieForge chooses the shape from your outcome. You care that the schedule and result are right, and that spend matches the job. See Controlling credits. Credit use for a run can appear on Executions when the platform records it.

​
What can go wrong