Mindee Receipt Extraction

arazzo: 1.0.1
info:
  title: Mindee Receipt Extraction
  summary: Enqueue a receipt with confidence scoring, poll until processed, then read the extracted line items and totals.
  description: >-
    Applies the Mindee asynchronous extraction pattern to expense receipts.
    The workflow uploads a receipt against a receipt extraction model with the
    confidence option enabled, polls the shared jobs endpoint until the job is
    Processed, and fetches the inference to read the supplier, totals, taxes,
    and line items along with per-field confidence levels useful for expense
    automation. 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: receipt-extraction
  summary: Upload a receipt with confidence scoring and read the extracted fields.
  description: >-
    Sends a receipt file to the extraction enqueue endpoint with confidence
    enabled, polls the job until processing finishes, and retrieves the parsed
    fields with their confidence attributes.
  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 receipts.
      file:
        type: string
        description: The receipt file to upload as binary form data.
      filename:
        type: string
        description: Optional filename to associate with the uploaded receipt.
  steps:
  - stepId: enqueueReceipt
    description: >-
      Send the receipt file to the asynchronous extraction queue with the
      confidence option enabled so each field carries a confidence level.
    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
        confidence: true
    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 receipt 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.enqueueReceipt.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 and their confidence attributes parsed from the receipt.
    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.enqueueReceipt.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      inferenceId: $response.body#/inference/id
      fields: $response.body#/inference/result/fields
  outputs:
    jobId: $steps.enqueueReceipt.outputs.jobId
    fields: $steps.getResult.outputs.fields