Mindee Invoice Extraction

arazzo: 1.0.1
info:
  title: Mindee Invoice Extraction
  summary: Enqueue an invoice for extraction, poll the job until processed, then read the extracted fields.
  description: >-
    The canonical Mindee asynchronous document AI pattern applied to invoices.
    The workflow uploads an invoice to the extraction queue against an
    extraction model, polls the shared jobs endpoint until the job status
    becomes Processed (or Failed), and then fetches the completed extraction
    inference to read the structured fields such as supplier, total, and dates.
    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
workflows:
- workflowId: invoice-extraction
  summary: Upload an invoice, wait for processing, and read the extracted fields.
  description: >-
    Sends an invoice file to the extraction enqueue endpoint, polls the job
    status until processing finishes, and retrieves the extraction inference
    result containing the parsed fields.
  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 extraction model trained for invoices.
      file:
        type: string
        description: The invoice file to upload as binary form data.
      filename:
        type: string
        description: Optional filename to associate with the uploaded invoice.
      alias:
        type: string
        description: Optional free-form identifier tagged onto the request.
  steps:
  - stepId: enqueueInvoice
    description: >-
      Send the invoice file to the asynchronous extraction queue using the
      invoice extraction model. Returns a job whose status begins as Processing.
    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.modelId
        file: $inputs.file
        filename: $inputs.filename
        alias: $inputs.alias
    successCriteria:
    - condition: $statusCode == 202
    outputs:
      jobId: $response.body#/job/id
      status: $response.body#/job/status
  - stepId: pollJob
    description: >-
      Poll the shared jobs endpoint for the enqueued invoice. The job remains
      Processing until extraction completes, then 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.enqueueInvoice.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 extraction inference and read the structured
      fields parsed from the invoice.
    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.enqueueInvoice.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      inferenceId: $response.body#/inference/id
      fields: $response.body#/inference/result/fields
  outputs:
    jobId: $steps.enqueueInvoice.outputs.jobId
    fields: $steps.getResult.outputs.fields