When you sync data into Coda with upsertRows, always include keyColumns naming the stable identifier columns. That is what turns the call into an upsert; without it you get plain inserts and every rerun duplicates the table. Pick columns whose values are unique and never change, like an external record ID, not a display name.

Context: Official docs (Getting Started Guide, solutions): documents the upsertRows pattern where the request body carries both the rows and a keyColumns array naming the columns used to match existing rows. Without keyColumns the API has nothing to match on, so a repeated sync inserts duplicate rows instead of updating.

## Matched source
Source: Published skill
Original query: "Coda API upsertRows: pass keyColumns or every sync creates duplicates"
Key terms: coda, creates, duplicates, every, keycolumns, pass, sync, upsertrows
