Mindee Split Multi-Document File

arazzo: 1.0.1
info:
  title: Mindee Split Multi-Document File
  summary: Enqueue a multi-document file for splitting, poll until processed, then read the identified document ranges.
  description: >-
    Uses Mindee's asynchronous split utility to identify the individual
    documents contained within a single multi-document file. The workflow
    uploads a file to the split queue, polls the shared jobs endpoint until the
    job is Processed, and fetches the split inference to read the list of
    document ranges. This is the natural first step when a scanned batch holds
    several documents that must be separated before extraction. 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: splitApi
  url: ../openapi/mindee-split-api-openapi.yml
  type: openapi
- name: jobsApi
  url: ../openapi/mindee-jobs-api-openapi.yml
  type: openapi
workflows:
- workflowId: split-multi-document
  summary: Upload a multi-document file and read the identified document ranges.
  description: >-
    Sends a file to the split enqueue endpoint, polls the job until processing
    finishes, and retrieves the split result containing the identified document
    ranges.
  inputs:
    type: object
    required:
    - authorization
    - modelId
    - file
    properties:
      authorization:
        type: string
        description: Mindee API key sent in the Authorization header.
      modelId:
        type: string
        description: UUID of the split utility model to apply.
      file:
        type: string
        description: The multi-document file to upload as binary form data.
      filename:
        type: string
        description: Optional filename to associate with the uploaded file.
  steps:
  - stepId: enqueueSplit
    description: >-
      Send the multi-document file to the asynchronous split queue. Returns a
      job whose status begins as Processing.
    operationId: Enqueue_Split_Product_Inference_v2_products_split_enqueue_post
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    requestBody:
      contentType: multipart/form-data
      payload:
        model_id: $inputs.modelId
        file: $inputs.file
        filename: $inputs.filename
    successCriteria:
    - condition: $statusCode == 202
    outputs:
      jobId: $response.body#/job/id
      status: $response.body#/job/status
  - stepId: pollJob
    description: >-
      Poll the shared jobs endpoint until the split job reports Processed or
      Failed.
    operationId: Get_Job_Status_v2_jobs__job_id__get
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: job_id
      in: path
      value: $steps.enqueueSplit.outputs.jobId
    - name: redirect
      in: query
      value: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $response.body#/job/status
    onSuccess:
    - name: jobProcessed
      type: goto
      stepId: getResult
      criteria:
      - context: $response.body
        condition: $.job.status == "Processed"
        type: jsonpath
    - name: jobPending
      type: goto
      stepId: pollJob
      criteria:
      - context: $response.body
        condition: $.job.status == "Processing"
        type: jsonpath
  - stepId: getResult
    description: >-
      Retrieve the completed split inference and read the list of documents
      identified within the source file.
    operationId: Get_Split_Product_Result_v2_products_split_results__inference_id__get
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: inference_id
      in: path
      value: $steps.enqueueSplit.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      inferenceId: $response.body#/inference/id
      splits: $response.body#/inference/result/splits
  outputs:
    jobId: $steps.enqueueSplit.outputs.jobId
    splits: $steps.getResult.outputs.splits