Create draft and patch document

​

Draft and publish workflow

1. PATCH /api/v2/documents/{documentId}/draft (this endpoint) — creates a new draft on the published document and applies your patch. Returns the new draftIdentifier.
2. (optional) PATCH /api/v2/documents/{documentId}/draft/{draftId} — apply further patches to that draft.
3. POST /api/v2/documents/{documentId}/draft/publish — publish the main draft live.

​

Addressing tiles (no miniUuid)

• New key → a new tile is created.
• Existing key, identical content (a verbatim round-trip of a GET) → treated as a no-op and dropped from the patch.
• Existing key, changed content → the tile is updated (the server regenerates its internal identifier).
• linked-type tabs always re-resolve — they are excluded from no-op detection.
• Query presentations are immutable server-side: replace a tile by key, don’t mutate it in place.
• model_extension_id - The query-model binding. This is read-only and managed by Omni.

​

AI summaries via linked tiles

{
  "queryPresentations": {
    "data": {
      "3": {
        "type": "linked",
        "name": "Summary",
        "sourceQueryPresentationKey": "2"
      }
    }
  }
}

​

Binding an upload as a Source tab

1. GET /api/v2/documents/{documentId} to read the document’s workbookModelId
2. POST /api/v1/uploads with modelId set to the draft’s workbook model (the connection decides warehouse staging vs S3-only)
3. PATCH /api/v2/documents/{documentId}/draft (this endpoint) with a csv or spreadsheet tab whose fileUploadId is the upload ID returned in step 2
4. POST /api/v2/documents/{documentId}/draft/publish to publish

{
  "queryPresentations": {
    "data": {
      "2": {
        "type": "csv",
        "name": "Customer data",
        "fileUploadId": "a1b2c3d4-5678-90ab-cdef-1234567890ab"
      }
    }
  }
}