arazzo: 1.0.1
info:
  title: Mindee Classify Then Extract
  summary: Classify an unknown document, then run extraction on the same file once the type is known, reading the parsed fields.
  description: >-
    A two-product routing pattern. The workflow first classifies an unknown
    document using the classification utility, waits for that job to finish and
    reads the predicted type, then enqueues the same file for extraction
    against the supplied extraction model, polls the extraction job, and reads
    the parsed fields. This mirrors a real intake pipeline where documents of
    unknown type are identified before the appropriate extraction model is
    applied. 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: classificationApi
  url: ../openapi/mindee-classification-api-openapi.yml
  type: openapi
- name: extractionApi
  url: ../openapi/mindee-extraction-api-openapi.yml
  type: openapi
- name: jobsApi
  url: ../openapi/mindee-jobs-api-openapi.yml
  type: openapi
workflows:
- workflowId: classify-then-extract
  summary: Classify a document, then extract structured fields from it.
  description: >-
    Classifies a document to discover its type, then enqueues the same file for
    extraction, polling each job to completion and reading the predicted type
    and the extracted fields.
  inputs:
    type: object
    required:
    - authorization
    - classificationModelId
    - extractionModelId
    - file
    properties:
      authorization:
        type: string
        description: Mindee API key sent in the Authorization header.
      classificationModelId:
        type: string
        description: UUID of the classification utility model.
      extractionModelId:
        type: string
        description: UUID of the extraction model to apply after classification.
      file:
        type: string
        description: The document file to upload as binary form data.
      filename:
        type: string
        description: Optional filename to associate with the uploaded document.
  steps:
  - stepId: enqueueClassification
    description: >-
      Send the document to the asynchronous classification queue to discover
      its type.
    operationId: Enqueue_Classification_Product_Inference_v2_products_classification_enqueue_post
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    requestBody:
      contentType: multipart/form-data
      payload:
        model_id: $inputs.classificationModelId
        file: $inputs.file
        filename: $inputs.filename
    successCriteria:
    - condition: $statusCode == 202
    outputs:
      classificationJobId: $response.body#/job/id
  - stepId: pollClassification
    description: >-
      Poll the shared jobs endpoint until the classification 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.enqueueClassification.outputs.classificationJobId
    - name: redirect
      in: query
      value: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $response.body#/job/status
    onSuccess:
    - name: classificationProcessed
      type: goto
      stepId: getClassification
      criteria:
      - context: $response.body
        condition: $.job.status == "Processed"
        type: jsonpath
    - name: classificationPending
      type: goto
      stepId: pollClassification
      criteria:
      - context: $response.body
        condition: $.job.status == "Processing"
        type: jsonpath
  - stepId: getClassification
    description: >-
      Read the predicted document type from the completed classification
      inference.
    operationId: Get_classification_Product_Result_v2_products_classification_results__inference_id__get
    parameters:
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: inference_id
      in: path
      value: $steps.enqueueClassification.outputs.classificationJobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      classification: $response.body#/inference/result/classification
  - stepId: enqueueExtraction
    description: >-
      Now that the document type is known, send the same file to the extraction
      queue against the chosen extraction model.
    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 document.
    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:
    classification: $steps.getClassification.outputs.classification
    fields: $steps.getExtraction.outputs.fields