Option 1: If your goal is to ensure that the table is not empty at all times. You can do the following
insert into a temp table. After the insert is successful (add a tail snap to the output view of the insert snap. The tail snap will wait for its input snap to complete before sending the last n documents to its output view)
use execute snap to truncate original table (using execute snap)
use execute snap to insert into original table as select * from temp table (using execute snap)
Option 2: If you are ok with truncating the table just before you start the insert.
use the execute snap to truncate table (enable an output view)
have the rest of the pipeline to read source data and insert into the table