Upsert generator

Upsert into Airtable: PATCH /records with performUpsert

Name your table, key and columns and get the statement Airtable actually accepts, with a guard so an older row never overwrites a newer one. Below it: how the mechanic works, what breaks, and how Datrise loads Airtable incrementally.

Generate the statement

Statement · PATCH /records with performUpsert

// PATCH https://api.airtable.com/v0/{baseId}/deals
// Authorization: Bearer {personal access token}
{
  "performUpsert": {
    "fieldsToMergeOn": ["id"]
  },
  "records": [
    {
      "fields": {
        "id": "…",
        "name": "…",
        "stage": "…",
        "amount": "…",
        "owner_id": "…",
        "updated_at": "…"
      }
    }
  ]
}

How the upsert works in Airtable

Airtable's REST API supports upserts directly: a PATCH to the records endpoint with performUpsert.fieldsToMergeOn matches incoming records to existing ones on up to three fields and updates or creates them. The response separates createdRecords from updatedRecords, which is the only way to tell an insert from an update. Fields you do not send are left alone with PATCH and cleared with PUT.

The limits shape the loader. Ten records per request and five requests per second per base means a 5,000-row entity takes 500 requests over at least 100 seconds, so send batches sequentially and back off on HTTP 429. The merge fields must identify one record; if two existing records match, the request fails rather than choosing. Land a focused field set, because bases also cap total records.

Before you run it

fieldsToMergeOn takes up to three fields and they must uniquely identify a record; when more than one existing record matches, the request fails rather than picking one.
Each request carries at most 10 records and each base allows 5 requests per second. A 5,000-row sync is 500 requests over at least 100 seconds; send them sequentially with backoff on HTTP 429.
The response lists createdRecords and updatedRecords ids separately, which is the only way to tell an insert from an update; a PATCH leaves unlisted fields untouched, a PUT clears them.

Questions people ask

Which fields can I merge on?

Up to three fields whose combination is unique in the table, typically the source record id stored in a single-line text field. Linked-record and formula fields cannot be merge fields.

How fast can I sync?

Five requests per second per base, ten records each, so about 50 records per second at best. Exceeding it returns 429 and a short lockout; throttle client-side.

PATCH or PUT?

PATCH for syncs: it updates the listed fields and leaves others untouched. PUT rewrites the record and blanks any field you omit, including ones people edit by hand in the base.

The same generator for other destinations

Skip writing the merge at all

Datrise lands CRM and SaaS entities into Airtable with this exact mechanic, a watermark on updated-at, and typed columns, so the statement above is what runs on your behalf. Join the waitlist to get early access.

Browse the integration catalog