just asking to confirm this. Apparently it seems that whenever you create a Python recipe, you must specify an output dataset within Dataiku. But this isn't always desirable when the python script is for example at the end of the datapipeline and injects the output to an external API.
It is already possible to create a Python recipe with no input dataset but not without an output dataset. I suppose this was by design, but would it be logical to make it available without an explicit required output dataset? I am now using the required dataset as a log.
It is structurally impossible to have recipes without any output because the concept of DSS is always to "build an output" rather than "run a recipe". Users who have needs similar to yours indeed create a "dummy" output dataset, where they either don't write anything or write "progress / status" kind of information.