I'm testing flows locally during development by simply typing "py .\flow.py" in powershell, and the results are inadvertently getting communicated up to my prefect cloud UI without an agent running. How is this possible?

Hey Jon, assuming you are running a flow using 2.0, agents and work queues are only required if you want to trigger flows via UI or API.

If you ran

prefect cloud login

or manually set

PREFECT_API_URL

and

PREFECT_API_KEY

, then local Prefect runs will send results to Prefect Cloud.

That's wild. So what's the best practice for separating my local test runs from say flows that have been promoted to DEV or PROD status? Is that where different named work queues come in? Or tags? Or can I even differentiate them in the UI based on which API key was used to login?

You can use profiles for that if you want to avoid sending anything to the Cloud account. For example:

prefect profile create dev

then:

prefect profile use dev

And you'll have a fresh profile that will only store results locally. Then you can switch back to the default profile or create a production profile you can use when you want to do things like deploy to production.

But what if I want my local runs and my dev and prod runs to all show up on the cloud ui, but be filterable? I see the cloud "flow run" webpage has filters for "deployment" and "tags" that seem useful for something like that

Ah - understood. In Organization accounts, separate workspaces for dev, production, etc. can be a good way to keep things separate. Otherwise, separate dev and production deployments are one option. Flow tags are another. Work queue tags are deprecated, but now flow run tags.

That code didn't work for me. Is it like this now?

from prefect import tags
with tags("db", "blue"):
    my_flow("foo")

Yes, I believe so - sorry, my mistake on that.