List A User AI Task's Answers

GraphQL endpoint: /user

Lists every run of one task, newest first — including runs still QUEUED or RUNNING, and answers superseded by a later run, so a document appears once per run rather than once in total. result and ranAt are null while a run is still going, and stay null on one that ended without reaching the model — error says why. prompt and outputShape are what was asked when the run was published, so an older answer stays interpretable after the task is edited.

Query

query ($taskId: ID!) {
  aiTaskRuns(taskId: $taskId) {
    records {
      documentId
      fileName
      status
      result
      error
      ranAt
      prompt
      outputShape
    }
    paginationInfo { nextCursor }
  }
}

Variables

{
  "taskId": "00000000-0000-4000-8000-000000000002"
}

Response

{
  "data": {
    "aiTaskRuns": {
      "paginationInfo": {
        "nextCursor": null
      },
      "records": [
        {
          "documentId": "00000000-0000-4000-8000-000000000001",
          "error": null,
          "fileName": "acme-jan.pdf",
          "outputShape": {
            "total": "number"
          },
          "prompt": "Extract the invoice details from this document.",
          "ranAt": "2026-01-15T09:00:00.000000Z",
          "result": {
            "total": 420
          },
          "status": "COMPLETED"
        }
      ]
    }
  }
}