The extraction workflow,
endpoint by endpoint.

Render the CleenUI Dropzone and provide its onFileUpload callback. The callback generates a stable UUID for each file, splits the file into binary chunks, sends every chunk with its absolute byte offset, and reports upload progress through updateFile.

const CHUNK_SIZE = 4 * 1024 * 1024; // 4 MB chunks

const stageFile = async (fileEntry, updateFile) => {
  const stagedFileCode = crypto.randomUUID();
  const chunks = [];

  for (let offset = 0; offset < fileEntry.file.size; offset += CHUNK_SIZE) {
    chunks.push({
      body: fileEntry.file.slice(offset, offset + CHUNK_SIZE),
      offset,
    });
  }

  await Promise.all(
    chunks.map(async ({ body, offset }, index) => {
      await mediaApi.stageMediaFileInChunks({
        fileId: stagedFileCode,
        offset,
        chunk: body,    // raw bytes · Content-Type: application/octet-stream
      });
      updateFile?.({ progress: (index + 1) / chunks.length });
    })
  );

  return { ...fileEntry, id: stagedFileCode };
};

// Wire to the CleenUI Dropzone component:
<Dropzone
  accept=".pdf,image/png,image/jpeg"
  onFileUpload={stageFile}
  showUploadedFiles
/>

After all files are staged, call this endpoint. It captures the temporary bytes, creates a persistent document batch, parses and encrypts each document in the background, stores successful files, and releases the staged in-memory bytes.

{
  "items": [
    {
      "extractionTypeCode": "YOUR_EXTRACTION_TYPE",
      "stagedFileCode": "78962f79-97d0-41a2-99ae-6eb758d13c5a",
      "password": null,
      "originalFilename": "TaxReturn2025.pdf",
      "hints": null
    }
  ]
}

{
  "batchId": "8a14dc31-c6cb-4ecd-a7bf-2e7d42530c14",
  "status": "Pending",
  "totalItems": 1
}

Poll this endpoint until the batch reaches a terminal state. A typical client polls every eight seconds, stops polling when the view is abandoned, and enforces a maximum attempt count or overall timeout.

{
  "batchId": "8a14dc31-c6cb-4ecd-a7bf-2e7d42530c14",
  "batchType": "UPLOAD_DOCUMENTS",
  "status": "Pending",
  "totalItems": 1,
  "completedItems": 0,
  "succeededItems": 0,
  "failedItems": 0,
  "createdUtc": "2026-07-15T01:00:28.1055312",
  "completedUtc": null,
  "items": [
    {
      "index": 0,
      "status": "Running",
      "blobId": null,
      "jobId": null,
      "errorCode": null,
      "errorMessage": null
    }
  ]
}

{
  "batchId": "8a14dc31-c6cb-4ecd-a7bf-2e7d42530c14",
  "batchType": "UPLOAD_DOCUMENTS",
  "status": "Completed",
  "totalItems": 1,
  "completedItems": 1,
  "succeededItems": 1,
  "failedItems": 0,
  "createdUtc": "2026-07-15T01:00:28.1055312",
  "completedUtc": "2026-07-15T01:00:30.0046745",
  "items": [
    {
      "index": 0,
      "status": "Completed",
      "blobId": 121387,
      "jobId": null,
      "errorCode": null,
      "errorMessage": null
    }
  ]
}

After the upload batch completes, submit the extraction jobs. Create one request item per successfully uploaded file and copy that item's blobId into extractionUploadBlobId.

[
  {
    "responseSetId": 1025,
    "extractionTypeCode": "YOUR_EXTRACTION_TYPE",
    "forEntityTypeId": 56,
    "forEntityId": 1025,
    "hints": null,
    "extractionUploadBlobId": 121387
  }
]

{
  "batchId": "7cb27722-3e91-4e8b-8468-ed632bf719c2",
  "status": "Pending",
  "totalItems": 1
}

Poll the same batch endpoint with the extraction batchId returned in step 4. For this batch, batchType is SUBMIT_JOBS. Use the same cadence as step 3 — every eight seconds, max 60 attempts. A successfully completed item will have a populated jobId; proceed to step 6 once the batch reaches a terminal state.

Once the extraction batch completes, call this endpoint once per succeeded item, passing the jobId as itemId. The response contains the extracted field values for that document. On the client, map the returned values to the question-response UI — each extracted field corresponds to a question on the form. When the same field is extracted with multiple candidate values, resolve duplicates by choosing the entry with the highest confidence score.

When the user clicks Confirm Values, submit all question-response values together with the extraction job ID. The endpoint persists the confirmed values to the response set and marks the extraction as complete.

{
  "extractionJobId": "7cb27722-3e91-4e8b-8468-ed632bf719c2",
  "responses": [
    {
      "questionId": 101,
      "value": "John Smith"
    },
    {
      "questionId": 102,
      "value": "2025-04-15"
    }
  ]
}

{
  "success": true
}