Deduplicate complete rows, or list key headers one per line to keep the first row for each selected value combination.
Inputs are limited to 1 MiB, 10,000 rows, and 256 columns.
Nothing is placed in a URL, account, or remote log.
Strict, bounded, local processing
An empty key selection compares every cell. A named selection defines identity from only those columns and keeps the first matching record.
With no key columns, two rows are duplicates only when every decoded cell matches. With keys, only those values form the comparison tuple.
The first occurrence is retained and later matches are removed without merging their other values.
A display name may not uniquely identify a person and an email can change. Use reviewed domain identifiers when deduplication affects an authoritative dataset.
