Complete Task Activity failing to link Storage Bucket Data (DOM/Extraction Results) in Document Understanding Flow

Help Document Understanding
1

Hi community,

I am encountering a critical issue when using the Complete Task activity to programmatically close a Document Validation Action.

The Workflow:

  1. Create Document Data Validation Action → Successfully creates the action.
  2. Assign Task → Successfully assigns the task to the user.
  3. Complete Task → Called within a logic branch to bypass manual validation when certain criteria are met.
    The Problem:
    I am struggling with the Task Data parameter of the Complete Task activity.
  • Scenario A: Leaving Task Data empty. The task moves to Completed, but when opening it in Action Center, I see the PDF file only. The fields are missing, and I get the error: “Fields could not be extracted. An error occurred while retrieving the extraction data!”
  • Scenario B: Passing JObject.Parse(extractionResults.Serialize). The activity fails or the resulting Action becomes corrupted. I receive the following error: “Unable to open action. The possible reason is the associated data was deleted when the action was processed. The required file cannot be found on storage bucket with id ‘15’ and path ‘…/output_results.zip’”

Technical Insight: It appears that the Complete Task activity expects a specific metadata wrapper to correctly map the ValidatedExtractionResults to the Storage Bucket’s output_results.zip. Without this, the Orchestrator doesn’t trigger the necessary backend synchronization between the uploaded DOM/Text files and the final validated state.

Questions:

  1. What is the exact JObject schema required for the Task Data property to ensure the Action remains viewable (with DOM and fields) in the “Completed” tab?
  2. Why does providing the serialized ExtractionResults cause a Storage Bucket file-not-found error, even though the input_results.zip exists?
  3. Is there a specific requirement to manually upload the output_results.zip to the Storage Bucket before calling Complete Task?
    image
    image432×241 12.5 KB
2

I ran into something similar, and from what I found the Task Data for Complete Task isn’t meant to be just the serialized ExtractionResults. The action expects the same wrapped payload structure Action Center generates, including references to the storage artifacts, so passing raw extraction JSON can break that mapping. Leaving it empty completes the task but skips the proper sync, which is why you only see the PDF and missing fields. I don’t think manually uploading output_results.zip is the intended fix, it should normally be created by the backend process, so the file-not-found error usually means the payload is referencing an artifact that was never generated or linked correctly.