Skip to content

Schedules & Periodic Tasks

In addition to one-off workflows, Simmate supports periodic tasks (schedules). These are useful for:

  • Regularly syncing data from third-party databases.
  • Performing weekly database maintenance.
  • Running "steady-state" evolutionary searches.

Starting the Scheduler

Simmate uses a single "scheduler" process to manage all periodic tasks. You only need to run one of these for your entire project.

simmate compute start-schedules

The scheduler process will:

  1. Search through all your installed apps for a schedules.py module.
  2. Register any tasks defined in those modules.
  3. Submit them to the queue database as WorkItems at their specified intervals.

Because the scheduler process simply delegates tasks by creating WorkItems, it never blocks. The actual execution of these tasks is picked up and handled asynchronously by your active Simmate Workers.

Creating a Schedule

To add a schedule to your own app, create a schedules.py file in your app directory and use the built-in @schedule decorator.

# my_app/schedules.py
from simmate.compute import schedule
from my_app.workflows import MyMaintenanceWorkflow

# Run every day at 10:30am
@schedule(interval="daily", at="10:30")
def run_daily_maintenance():
    MyMaintenanceWorkflow.run()

# Run every hour
@schedule(interval="hourly")
def check_for_updates():
    print("Checking for updates...")

# Run every Saturday at 1:00am
@schedule(interval="weekly", on="saturday", at="01:00")
def run_weekly_cleanup():
    print("Running weekly cleanup...")

As long as my_app is in your settings.yaml under apps, the start-schedules command will find and register these tasks automatically.

Best Practices

  • Workers Required: Since the SimmateScheduler does not run the tasks itself, you must ensure you have at least one worker running (simmate compute start-worker) to process the scheduled jobs from the database queue.
  • Task Delegation: Decorating a task ensures that whenever the schedule interval hits, the scheduler submits the job directly as a WorkItem. Heavy lifting is naturally offloaded, preventing the scheduler loop from ever clogging.