@Marvin what’s the difference between prefect.runtime.get_tags() and prefect.context.TagsContext()

Here's a breakdown of the differences between

prefect.runtime.get_tags()

and

prefect.context.TagsContext()

based on the available documentation: 1. Prefect Runtime (

prefect.runtime.get_tags()

): - This is a part of the Prefect runtime tools that provides access to metadata and information about the current flow or task run. - Although I couldn't find a direct reference to

prefect.runtime.get_tags()

, the general use of runtime context like

prefect.runtime

is to retrieve real-time information related to flows and tasks that are executing. 2. Prefect Context (

prefect.context.TagsContext()

): -

prefect.context

is used to access contextual information during flow and task runs. It provides the execution context for tasks and flows. - The runtime API context allows access to attributes related to the current run, and using functions like

prefect.context.get_run_context()

provides direct access to the run context. -

TagsContext()

isn't directly mentioned either, but it could be a pattern to create or manage tags within the context. Prefect often uses contexts to encapsulate settings and states that are active during a run. The functionalities of these methods might have evolved between different Prefect versions (2.x and 3.x), but the documentation I accessed didn't give specifics on these exact methods. It's possible these functions or classes are used internally or have recently been introduced. For additional details, you might want to check the Prefect runtime context documentation for Prefect 2.x and Prefect 3.x.

tldr the runtime module is explicitly for user consumption, the

TagsContext

is an implementation detail that you wont find in the docs

is it possible for the two to return different values? i.e. will this ever fail the assertion:

from prefect.runtime import flow_run
from prefect.context import TagsContext

assert flow_run.get_tags() == TagsContext.get().current_tags

yes - I would recommend against using

TagsContext

directly, because when you do, you're not necessarily talking about the tags context that the flow run is using

we’ve been using tags to fork logic in different flows/tasks and noticed the TagsContext.get().current_tags doesn’t always carry things downstream; though it works when I run the flow locally

hmm im still not sure why you need them from both places, sorry could be missing something

In [1]: from prefect.context import get_run_context

In [2]: from prefect import task

In [3]: @task(tags={"a", "b", "c"})
   ...: def t():
   ...:     print(get_run_context().task_run.tags)
   ...:

In [4]: t()
['a', 'b', 'c']

this is how I would access the tags for a current flow/task

ah I see; we definitely don’t need both, just wanted to know which is the correct one

👍 yeah

runtime

and

get_run_context

are both "public" interfaces and more or less providing access to the same things

runtime

is just convenient for stuff like

task_run_name

or

result_storage_key

so you can template extra stuff into those things