@Marvin I have a job that is scheduled from the Prefect UI and it doesn’t take parameters. This job generates a value internally (a snapshot_date). I want to pass this value to the next job using automations. Right now I’m considering embedding the value in the flow run name and then using Jinja templating + string parsing to extract it for downstream jobs, but that feels fragile. What’s the best pattern in Prefect for passing internally-generated values between scheduled jobs when parameters aren’t available? Should I be using tags, artifacts, or some other approach?

✅ thought for 596.0 seconds

You can any time emit custom events in your flow runs that hold custom data. Based on this you can trigger other automations.

You’re right to be wary of encoding data in the flow run name — that’s brittle and hard to evolve. In Prefect 3.x, the two most robust patterns for passing a runtime value like

snapshot_date

between flows scheduled independently (via Automations) are: 1) Event → Automation → RunDeployment with parameters (preferred) - Have the producer flow emit an event that includes the value in its payload. - Create an Automation that triggers on that event and runs the downstream deployment, templating the value from the event into the downstream’s parameters. Producer flow:

from prefect import flow
from prefect.events.utilities import emit_event

@flow
def producer():
    # compute your value
    snapshot_date = "2025-10-01"
    # emit a custom event with the value in the payload
    emit_event(
        event="myapp.snapshot.created",
        resource={"prefect.resource.id": f"snapshot:{snapshot_date}"},
        payload={"snapshot_date": snapshot_date},
    )

Automation (UI): - Trigger: Event trigger on

myapp.snapshot.created

- Action: Run Deployment - Parameters (JSON):

{
  "snapshot_date": "{{ event.payload.snapshot_date }}"
}

Downstream flow:

from prefect import flow

@flow
def consumer(snapshot_date: str):
    # use snapshot_date
    ...

Notes and links: - This keeps your value explicit, typed, and auditable via events. - Events utilities: Event-driven flows with Prefect - Artifacts also emit events you can trigger on if you prefer that surface: Artifacts API 2) If you truly cannot add a parameter to the downstream: pass via job variables - RunDeployment actions can set job variables. - The downstream flow can read them at runtime via

prefect.runtime.flow_run.job_variables

. Automation (UI): - Trigger: Same as above (your custom event or an

artifact.created

if you go the artifact route) - Action: Run Deployment - Job variables (JSON):

{
  "SNAPSHOT_DATE": "{{ event.payload.snapshot_date }}"
}

Downstream flow:

from prefect import flow
from prefect.runtime import flow_run

@flow
def consumer():
    snapshot_date = flow_run.job_variables.get("SNAPSHOT_DATE")
    ...

Other options and tradeoffs - Artifacts: You can create an artifact containing the value and trigger on

artifact.created

, then template fields from the event into parameters/job variables. This works well if you already publish artifacts, but you’ll want to ensure the artifact’s event payload reliably contains the value you need. API: Artifacts API - Variables/Blocks: Persist to a “latest” location and have the downstream read from it (e.g., write

latest_snapshot_date

to a custom Block or external store). This is simple but “last write wins” can be racy if multiple producers run concurrently; it’s also less traceable than the event approach. Blocks overview: Blocks API What I’d recommend - If you can, add an optional

snapshot_date

parameter to the downstream and use the event → automation → parameter pattern. It’s the cleanest and most maintainable. - If you can’t modify the downstream signature, use job variables as a transport and read them via

prefect.runtime.flow_run.job_variables

in the flow. Quick question to tailor this: Are you able to update the downstream flow to accept an optional parameter? And do you prefer emitting a small custom event from the producer or reusing an artifact-created event from something you already generate?