arazzo: 1.0.1
info:
  title: Nanonets Sync Predict and Review
  summary: Run a sync OCR prediction on a small file, then approve or hold it for review.
  description: >-
    A human-in-the-loop extraction flow for files of three pages or fewer. The
    workflow runs a synchronous OCR prediction on an uploaded file, then branches
    on the returned page: when a prediction is present it marks the file approved
    (moderated), and when no prediction came back it leaves the file unapproved
    for manual review. 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
- name: fileManagementApi
  url: ../openapi/nanonets-file-management-api-openapi.yml
  type: openapi
workflows:
- workflowId: sync-predict-and-review
  summary: Synchronously predict on a file then approve or hold it based on the result.
  description: >-
    Runs a sync OCR prediction on a file and branches to approve the file when a
    prediction is returned or to unapprove it for review when none is.
  inputs:
    type: object
    required:
    - authorization
    - modelId
    - file
    properties:
      authorization:
        type: string
        description: HTTP Basic credential header value (Basic <base64 of apiKey:>).
      modelId:
        type: string
        description: Unique identifier for the Nanonets OCR model.
      file:
        type: string
        description: Binary contents of the document to run sync OCR against.
      requestMetadata:
        type: string
        description: Free-form identifier echoed back in the prediction response.
  steps:
  - stepId: predict
    description: Run a synchronous OCR prediction on the uploaded file.
    operationId: ocrModelLabelFileByModelIdPost
    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
        request_metadata: $inputs.requestMetadata
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      requestFileId: $response.body#/result/0/request_file_id
      predictions: $response.body#/result/0/prediction
    onSuccess:
    - name: hasPrediction
      type: goto
      stepId: approveFile
      criteria:
      - context: $response.body
        condition: $.result[0].prediction.length > 0
        type: jsonpath
    - name: noPrediction
      type: goto
      stepId: holdForReview
      criteria:
      - context: $response.body
        condition: $.result[0].prediction.length == 0
        type: jsonpath
  - stepId: approveFile
    description: Mark the predicted file as approved (moderated).
    operationId: verifyFile
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: model_id
      in: path
      value: $inputs.modelId
    - name: request_file_id
      in: path
      value: $steps.predict.outputs.requestFileId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      approveResponse: $response.body
    onSuccess:
    - name: done
      type: end
  - stepId: holdForReview
    description: >-
      Leave the file in the unapproved state so a reviewer can inspect it before
      it is moderated.
    operationId: unverifyFile
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: model_id
      in: path
      value: $inputs.modelId
    - name: request_file_id
      in: path
      value: $steps.predict.outputs.requestFileId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      holdResponse: $response.body
  outputs:
    requestFileId: $steps.predict.outputs.requestFileId
    approveResponse: $steps.approveFile.outputs.approveResponse
    holdResponse: $steps.holdForReview.outputs.holdResponse