Yes — in Prefect 3.x you can pass values from the triggering event into a Run Deployment action using Jinja templates. How to reference your event payload - In an Automation action, parameters can be templated with

{{ event.payload.<key> }}

. - For your example, use

{{ event.payload.table_name }}

. If you’re configuring the Automation in the UI - Add an action: Run deployment - In Parameters, add a key

table_name

- Click the template toggle (</>) and set the value to:

{{ event.payload.table_name }}

If you’re defining the Automation as JSON (or via API), templated fields use the special JSON form:

"parameters": {
  "table_name": {
    "__prefect_kind": "jinja",
    "template": "{{ event.payload.table_name }}"
  }
}

Notes - Make sure your deployment’s flow has a

table_name

parameter in its signature. - You can add simple logic/defaults in the template, e.g.:

{{ event.payload.table_name or 'unknown_table' }}

Docs - Using templates in Automations - Available Jinja filters and context If you share how you’re creating the Automation (UI vs code) I can provide the exact steps or a ready-to-paste snippet.

Absolutely — you can wire this up in the UI so your automation passes

table_name

from the event payload to your deployment. Here are the UI steps in Prefect 3.x: 1) Open Automations - In the Prefect UI, go to Automations. - Click “New automation”. 2) Name and Trigger - Give it a name like “Update dashboard widget on gold table update”. - Trigger: choose “Event occurs”. - Condition: set Event name equals

axos.gold.table.updated

. 3) Action: Run deployment - Add action: “Run deployment”. - Select your deployment: “dashboard-widget-update” (or the exact name shown in your Deployments list). - Parameters: - Add a parameter key:

table_name

- Click the template toggle (</>) for the value field - Enter:

{{ event.payload.table_name }}

- Save the action. 4) Save and test - Save the automation. - When the event you showed is emitted, the automation will run the deployment and pass the value

gold.invoice_monthly_summary

into your flow’s

table_name

argument (matching your flow signature

def update_dashboard_widget(table_name):

). Notes and options: - Your event also includes

resource.table

. If you prefer that source, you can use:

{{ event.resource.table }}

- Make sure your deployment’s flow parameter is named exactly

table_name

. If it differs, update either the flow signature or the parameter key in the action. Docs - Using templates in Automations - Available Jinja filters/context If you share a screenshot of your deployment’s parameters panel, I can double-check the exact field names and confirm the template entry for you.

Pick “Use form input and Jinja input”. Then: - Click Add parameter - Key:

table_name

- Click the template toggle (</>) for the value - Paste:

{{ event.payload.table_name }}

- Save If you prefer “Use JSON input”, the equivalent is:

{
  "table_name": {
    "__prefect_kind": "jinja",
    "template": "{{ event.payload.table_name }}"
  }
}

Docs: - Using templates in Automations - Available Jinja filters/context

Great question. Short answer: in Prefect 3.x there isn’t a CLI command to create automations, so you won’t pass a

prefect automation create ...

command to your subprocess runner. Instead, create the automation via Python using the Prefect client. Below is a minimal async script that: - looks up your deployment by name (

update-dashboard-widget/dashboard-widget-update

) - creates an event-triggered automation with your Jinja parameter mapping for

table_name

- sets the action to “Run deployment” with empty job variables You can call this from your “run once” script.

import asyncio
from prefect.client.orchestration import get_client
from prefect.events.schemas.automations import AutomationCore, EventTrigger, Posture
from prefect.events.actions import RunDeployment

FLOW_NAME = "update-dashboard-widget"
DEPLOYMENT_NAME = "dashboard-widget-update"
AUTOMATION_NAME = "update dashboard trigger"

async def create_automation():
    async with get_client() as client:
        # Resolve deployment id by flow and deployment name
        deployment = await client.read_deployment_by_name(
            name=DEPLOYMENT_NAME,
            flow_name=FLOW_NAME,
        )
        if deployment is None:
            raise RuntimeError(f"Deployment not found: {FLOW_NAME}/{DEPLOYMENT_NAME}")

        # Build the automation definition
        automation = AutomationCore(
            name=AUTOMATION_NAME,
            enabled=True,
            trigger=EventTrigger(
                match={},
                match_related={},
                after=[],
                expect=["axos.gold.table.updated"],
                for_each=[],
                posture=Posture.Reactive,
                threshold=1,
                within=10,
            ),
            actions=[
                RunDeployment(
                    deployment_id=deployment.id,
                    parameters={
                        "table_name": {
                            "__prefect_kind": "jinja",
                            "template": "{{ event.payload.table_name }}",
                        }
                    },
                    job_variables={},
                )
            ],
        )

        created = await client.create_automation(automation)
        print(f"Created automation: {created.name} ({created.id})")

if __name__ == "__main__":
    asyncio.run(create_automation())

Notes: - This assumes your script is already authenticated to Prefect Cloud (via

PREFECT_API_KEY

and

PREFECT_API_URL

) and the correct workspace is set. - If you want the action to reference the deployment by name instead of id, we can adjust the payload, but using

deployment_id

is the most reliable. - The trigger body mirrors what you pasted:

type: "event"

,

posture: "Reactive"

,

expect: ["axos.gold.table.updated"]

, etc. If you really want to stick with your

subprocess.run(...)

approach, you could wrap this Python in a standalone file and call

python path/to/create_automation.py

from your script. There isn’t a supported

prefect automation create

CLI in 3.x today, so Python is the recommended route. Would you like me to add existence-check/update logic so the script is idempotent (create if missing, update if present)?