```jobby_job_prod = AssetsDefinition.from_graph( ...
# ask-community
j
Copy code
jobby_job_prod = AssetsDefinition.from_graph(
    jobby_job_graph,
    partitions_def=HourlyPartitionsDefinition(start_date=PARTITION_START_DATE),
    keys_by_output_name={"result": AssetKey(["augmentation_data", "i_metadata"])},
)
how do i define non_argument_deps for AssetsDefinition.from_graph ? (edited) there’s an upstream dep that’s not an argument passed in
non_argument_deps
. Isn’t an option in
AssetsDefinition.from_graph
o
hi @Jeremy! there isn't a way to do this directly from the
from_graph()
call at the moment. Under the hood, the
@asset
decorator is creating a new Op, with Nothing-type inputs corresponding to each of the
non_argument_deps
. Because the
from_graph
method works on existing graphs (rather than creating new ones), it doesn't do this step. To work around it, one solution would be to add a new Nothing input to the ip_augmentation_graph and wire it to the first op in the graph, i.e.:
Copy code
# before ***************

@op
def first_op(x):
    ...

@graph
def graph_asset(regular_asset):
    return second_op(first_op(regular_asset))

# after ****************

@op(ins={"x": In(), "start_after": In(Nothing)})
def first_op(x):
    ...

@graph
def graph_asset(reg_asset, non_arg_asset):
    return second_op(first_op(regular_asset, start_after=non_arg_asset))