Nanonets Async OCR Predict and Poll

arazzo: 1.0.1
info:
  title: Nanonets Async OCR Predict and Poll
  summary: Submit a large document for async OCR, then poll until the file-level prediction is ready.
  description: >-
    The recommended pattern for documents larger than three pages. The workflow
    uploads a local file to a Nanonets OCR model in async mode, captures the
    returned request_file_id, then repeatedly fetches the file-level inference
    until a prediction is available, branching back to poll again while the
    result is still pending. Every step spells out its request inline so the flow
    can be read and executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: ocrApi
  url: ../openapi/nanonets-ocr-api-openapi.yml
  type: openapi
workflows:
- workflowId: async-ocr-predict-and-poll
  summary: Async predict on a file then poll the inference request for the finished result.
  description: >-
    Uploads a file to a Nanonets OCR model in async mode and polls the inference
    request endpoint until the file prediction is returned.
  inputs:
    type: object
    required:
    - authorization
    - modelId
    - file
    properties:
      authorization:
        type: string
        description: >-
          HTTP Basic credential header value (Basic <base64 of apiKey:>) where the
          Nanonets API key is the username and the password is empty.
      modelId:
        type: string
        description: Unique identifier for the Nanonets OCR model.
      file:
        type: string
        description: Binary contents of the document to run async OCR against.
      requestMetadata:
        type: string
        description: Free-form identifier echoed back in the prediction response.
  steps:
  - stepId: submitAsync
    description: >-
      Upload the file to the model in async mode and capture the request_file_id
      used to poll for the result.
    operationId: ocrModelLabelFileAsyncByModelIdPost
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: model_id
      in: path
      value: $inputs.modelId
    requestBody:
      contentType: multipart/form-data
      payload:
        file: $inputs.file
        async: true
        request_metadata: $inputs.requestMetadata
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      requestFileId: $response.body#/result/0/request_file_id
      submitStatus: $response.body#/result/0/status
  - stepId: pollPrediction
    description: >-
      Fetch the file-level prediction for the request_file_id. Branch back to
      poll again while no prediction array is present, otherwise finish.
    operationId: ocrModelGetPredictionFileByFileId
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: model_id
      in: path
      value: $inputs.modelId
    - name: request_file_id
      in: path
      value: $steps.submitAsync.outputs.requestFileId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      moderatedCount: $response.body#/moderated_images_count
      unmoderatedCount: $response.body#/unmoderated_images_count
      signedUrls: $response.body#/signed_urls
    onSuccess:
    - name: stillPending
      type: goto
      stepId: pollPrediction
      criteria:
      - context: $response.body
        condition: $.moderated_images_count == 0 && $.unmoderated_images_count == 0
        type: jsonpath
    - name: ready
      type: end
  outputs:
    requestFileId: $steps.submitAsync.outputs.requestFileId
    moderatedCount: $steps.pollPrediction.outputs.moderatedCount
    signedUrls: $steps.pollPrediction.outputs.signedUrls