date_histogram is not support as aggregation in transform yet but only as part of
group_by. In your config in the screenshot you specify date_histogram twice, however the definition is
group_by is all you need.
More problematic regarding support is
serial_diff, which is a pipeline aggregation. Pipeline aggregations do not fit with the paging logic we internally use in transform. The necessary previous bucket might be in the previous page and therefore inaccessible for the current page. In this usecase this might be solvable, but for other pipeline aggregation its not that easy. We might work on a solution for this, but I do not see this anytime soon, because of complex problems.