Hi! I've shown up in this community today because I'm starting to evaluate Prefect against Dagster for building some brand new workflows as part of a greenfield project at my company. We're primarily looking at scaling out web scraping jobs and using LLMs+heuristics to capture structured information to power certain product features. I have a kind of annoying question out of the gate. I prefer to use Pyright (my type checker of choice) in strict mode whenever possible, and when I tried the example code from the get-started guide with this, Pyright complained about

from prefect import flow, task

because neither

flow

nor

task

are fully typed; they have some parameters where the type definition is partially unknown. So: does Prefect type check its code against strict-mode checkers on deployment, and is there perhaps something I've gotten wrong in my config? Or is strict mode checking not going to be realistic for my Prefect projects? (to be fair, I had the same thing happen with Dagster, so...)

hi @Nash Taylor - as far as type checking / coherence, this is something we’re actively trying to improve. there are a couple tricky problems around sync/async interfaces that we need to spend some design time on, but in general if you find your IDE complaining about something simply underspecified, it’d be super helpful for you to open an issue (or PR :)) so it’s on our radar and we can fix it.

Where a task/flow signature contains a pydantic model as a type, are those arguments runtime validated? (Basically

pydantic.validate_call

)?

yes! even if it’s not a pydantic type