Mindee Split Then Extract

arazzo: 1.0.1
info:
  title: Mindee Split Then Extract
  summary: Split a multi-document file into ranges, then extract structured fields from the original file, reading splits and fields.
  description: >-
    A two-product batch intake pattern. The workflow first runs the split
    utility to discover the individual document ranges inside a multi-document
    file, waits for that job and reads the ranges, then enqueues the same file
    for extraction against the supplied extraction model, polls the extraction
    job, and reads the parsed fields. This mirrors a scanning pipeline where a
    batch file is segmented before its contents are read. 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: extractionApi
  url: ../openapi/mindee-extraction-api-openapi.yml
  type: openapi
- name: jobsApi
  url: ../openapi/mindee-jobs-api-openapi.yml
  type: openapi
- name: splitApi
  url: ../openapi/mindee-split-api-openapi.yml
  type: openapi
workflows:
- workflowId: split-then-extract
  summary: Split a multi-document file, then extract fields from it.
  description: >-
    Splits a batch file into document ranges, then enqueues the same file for
    extraction, polling each job to completion and reading the ranges and the
    extracted fields.
  inputs:
    type: object
    required:
    - authorization
    - splitModelId
    - extractionModelId
    - file
    properties:
      authorization:
        type: string
        description: Mindee API key sent in the Authorization header.
      splitModelId:
        type: string
        description: UUID of the split utility model.
      extractionModelId:
        type: string
        description: UUID of the extraction model to apply after splitting.
      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 to discover
      its document ranges.
    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.splitModelId
        file: $inputs.file
        filename: $inputs.filename
    successCriteria:
    - condition: $statusCode == 202
    outputs:
      splitJobId: $response.body#/job/id
  - stepId: pollSplit
    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.splitJobId
    - name: redirect
      in: query
      value: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $response.body#/job/status
    onSuccess:
    - name: splitProcessed
      type: goto
      stepId: getSplit
      criteria:
      - context: $response.body
        condition: $.job.status == "Processed"
        type: jsonpath
    - name: splitPending
      type: goto
      stepId: pollSplit
      criteria:
      - context: $response.body
        condition: $.job.status == "Processing"
        type: jsonpath
  - stepId: getSplit
    description: >-
      Read the identified document ranges from the completed split inference.
    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.splitJobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      splits: $response.body#/inference/result/splits
  - stepId: enqueueExtraction
    description: >-
      Send the same file to the extraction queue against the chosen extraction
      model to read the contents of the documents.
    operationId: Enqueue_Extraction_Product_Inference_v2_products_extraction_enqueue_post
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    requestBody:
      contentType: multipart/form-data
      payload:
        model_id: $inputs.extractionModelId
        file: $inputs.file
        filename: $inputs.filename
    successCriteria:
    - condition: $statusCode == 202
    outputs:
      extractionJobId: $response.body#/job/id
  - stepId: pollExtraction
    description: >-
      Poll the shared jobs endpoint until the extraction 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.enqueueExtraction.outputs.extractionJobId
    - name: redirect
      in: query
      value: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $response.body#/job/status
    onSuccess:
    - name: extractionProcessed
      type: goto
      stepId: getExtraction
      criteria:
      - context: $response.body
        condition: $.job.status == "Processed"
        type: jsonpath
    - name: extractionPending
      type: goto
      stepId: pollExtraction
      criteria:
      - context: $response.body
        condition: $.job.status == "Processing"
        type: jsonpath
  - stepId: getExtraction
    description: >-
      Retrieve the completed extraction inference and read the structured
      fields parsed from the documents.
    operationId: Get_Extraction_Product_Result_v2_products_extraction_results__inference_id__get
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: inference_id
      in: path
      value: $steps.enqueueExtraction.outputs.extractionJobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      inferenceId: $response.body#/inference/id
      fields: $response.body#/inference/result/fields
  outputs:
    splits: $steps.getSplit.outputs.splits
    fields: $steps.getExtraction.outputs.fields