Yes, you can define a Python recipe with an ouput dataset but no input. Use case could be that you retrieve data from an external API, process it in a Pandas dataframe, and then save it to your output dataset using a dataiku method such as write_with_schema.
Regarding this particular function (scenario.build_dataset()), is there a list of accepted parameters for the build_mode keyword argument? I saw that RECURSIVE_BUILD is the default parameter, but is there something similar for non-recursive ones?
Here are the available options: /** Rebuild what is required for dependencies */ RECURSIVE_BUILD, /** Only rebuild the dataset directly, ignore the state of the dependencies */ NON_RECURSIVE_FORCED_BUILD, /** Rebuild all recursively, ignore the state of the dependencies */ RECURSIVE_FORCED_BUILD, /** Recursive build, but only build "missing" datasets, don't refresh out of date ones */ RECURSIVE_MISSING_ONLY_BUILD