How can I combine (in the experimental software defined assets api) steps with regular operations? I.e. have a foreign asset as the input perhaps some dynamic graph/ops in between and then an asset per final output of the dynamic graph (or a fan-in operation and a single asset)?

Hi, we don't support this functionality at the moment, though we'll be working on asset and op interoperability very soon!

do you already have an issue to track for this topic?

Unfortunately there is no traceable issue at the moment, but asset-op interoperability will definitely be a work in progress over the next several months.

Hi @claire - did this end up being implemented? I want to "tag on" an op at the end of a DBT / python / DBT job to do some analysis and log some stuff to MLFlow. This doesn't necessarily fit the SDA pattern as it is just performing some calculations in python and logging to MLFLOW but it uses an asset as input. This looks like it might fit my usage but feels a bit heavy to define a single op based graph (with no output) and convert to an asset.

Hi Martin. Unfortunately there isn't a great built in way to run an op that accepts an asset input. You could either do the graph-backed asset approach you linked above (you'd still have to define an output as an asset), or you could simply just have a downstream asset that performs your operation and outputs a null value.