With the Google Cloud command line CLI running you can specify a local jar with the --jars flag. However I want to submit a job using the Python API. I have that working but when I specify the jar, if I use the file: prefix, it looks on the Dataproc master cluster rather than on my local workstation.
There is an easy workaround which is to just upload the jar using the GCS library first but I wanted to check if the Dataproc client libraries already supported this convenience feature.
Not at the moment. As you mentioned, the most convenient way to do this strictly using the Python client libraries would be to use the GCS client first and then point to your job file in GCS.