It's not possible anymore to create a ML workflow on a full corpus

The backend gives a 400 complaining about a missing element.