I have a graph which sometimes I want to run by itself, and sometimes as part of a different graph. It was all working fine by itself. I just decorated the function that defines the graph with

@graph

, created a job via

to_job

and imported the job in my

repo.py

. In order to run it as part of another graph I added

ins

and

out

to the

@graph

decorator. That also runs fine. However since I added the

ins

and

out

to the

@graph

I can’t launch dagit anymore. It’s the job described above that’s causing the problem. it either complains about

Missing required config entry "inputs" at the root

or when I add

inputs

to my yaml it complains that

Error 1: Received unexpected config entry "inputs" at the root.

question: do I have to define 2 separate graphs in this case?

are the

inputs

that are mapped on the graph intended to be optional? If yes, you can provide them with a default_value which I believe will resolve the issue. This can be done on the

@op

decorated function like you would normally set default values for function inputs.

hm no they’re not optional. they are required by the first op within the graph

maybe this is useful: https://docs.dagster.io/concepts/ops-jobs-graphs/jobs-graphs#conditional-branching

I filled out the details below

inputs

. yes the error goes on

Error 1: Received unexpected config entry "inputs" at the root. Expected: "{ index_and_match: { config: {

and then what follows are all the other ops of the graph with their configs/inputs

The structure of the config will be different when the graph is used as subgraph are you trying to provide the inputs information the same way you were when the job was directly using the graph?

ah that could be it. yes i am still reading the same yaml file which has

ops

at the root. what’s the structure of the config when the graph takes inputs?

graph

at the root?

I forget the exact structure off the top of my head but

dagit

should ideally display it in a way that makes it clear

ah but the problem is that dagit doesn’t launch because of this 😄

Ah you are setting this on the job directly with

config=

? I would comment that out - get the right thing built in dagit then go back and add it corrected.

Created issue at: https://github.com/dagster-io/dagster/issues/6440

ah very nice,

inputs

just have to be defined at the root level (same level as

ops

). thanks!