What you would like to accomplish:

https://cloud.google.com/ai-platform/prediction/docs/reference/rest/v1/projects.jobs/create

Requires a Job ID when creating jobs. Calls to jobs.create with a re-used ID result in errors.

This complicates automated job creation, as clients must be able to generate dynamic requests (for unique Job IDs). A change to the ML API would allow for simpler configurations in Cloud Scheduler or other automation solutions.

How this might work:

Without removing the current behavior, update the API for ml.projects.jobs.create to optionally allow for automatic JobID creation:

• Add a new field to the job object named "JobName"
• Add another new field specifying that this request requires automatic Job ID generation
• The API should create the Job with JobID=JobName+generated_UUID and return the generated Job ID

If applicable, reasons why alternative solutions are not sufficient:

In the context of a solution like Cloud Scheduler, which cannot generate dynamic requests, an intermediary such as GCF is needed to generate requests that create jobs reliably. This adds complexity when trying to schedule ML jobs with GCP Native solutions.