Configuration¶

everystep has no settings module and reads no environment variables. It is configured in three places:

Your Django settings — the app and the database.
The worker command line — everything a runner does.
Optional extras — observability integrations, enabled by installing a package.

Django¶

Add "everystep" to INSTALLED_APPS and run migrations (see the quickstart).

The library itself works on any database backend your project uses: scheduling and the models are plain Django. Workers run on PostgreSQL and MariaDB. On PostgreSQL, claiming uses SELECT ... FOR UPDATE SKIP LOCKED, so concurrent workers claim disjoint batches without blocking. MariaDB has no SKIP LOCKED, so the claim falls back to a plain FOR UPDATE: claims stay exclusive but serialize while a batch is being locked.

Worker flags¶

python manage.py everystep_worker [flags]:

Flag	Default	Meaning
`--pool N`	`4`	Thread pool size: the maximum number of workflows this worker runs at once.
`--poll S`	`0.2`	Seconds between claim polls. Bounds how early a due workflow can be picked up.
`--name NAME`	hostname	Runner identity: recorded on claimed runs, used by the UI and metrics. Unique among concurrently running workers; keep it identical across restarts so a restart reclaims the runs a crash left behind. See workers.
`--drain S`	`30`	Seconds to wait for in-flight steps after SIGTERM/SIGINT, at step boundaries, before requeueing the runs and exiting. `0` waits indefinitely. See rollouts.
`--metrics-port N`	`0` (off)	Serve this worker's Prometheus metrics on the given TCP port. Requires the `metrics` extra.
`--metrics-bind ADDR`	`0.0.0.0`	Interface to bind the metrics endpoint to (in-cluster scraping).

Optional extras¶

All three degrade to no-ops when not installed: every call site checks for the package before doing anything.

Extra	Install	Enables
`sentry`	`pip install "everystep[sentry]"`	Unhandled workflow failures reported to Sentry, with the workflow's name and id attached. See Sentry.
`metrics`	`pip install "everystep[metrics]"`	Prometheus metrics, the per-runner metrics endpoint, and the host-application metrics view. See metrics.
`otel`	`pip install "everystep[otel]"`	OpenTelemetry spans for runs and executed steps on the global tracer `everystep`. See traces.

URL mounts¶

Nothing is mounted for you. Add what you need:

# urls.py
from django.urls import include, path
from everystep import views

urlpatterns = [
    path("everystep/", include("everystep.urls")),          # the UI
    path("everystep/metrics", views.metrics_view),     # Prometheus endpoint with DB-derived queue gauges
]

The UI is optional; so is the metrics view (without it, in-process everystep metrics are still served from your existing /metrics, and the per-runner endpoint is available via --metrics-port). Both endpoints are unauthenticated — keep them behind network segmentation or your own authentication.

Logging¶

Worker events (requeued workflows, runs that crashed outside the runner) are logged through the everystep logger. Configure it like any other in your application; there are no everystep-specific logging settings.