Mindee Extract From URL

arazzo: 1.0.1
info:
  title: Mindee Extract From URL
  summary: Enqueue a publicly hosted document by URL for extraction, poll until processed, then read the extracted fields.
  description: >-
    A variant of the Mindee asynchronous extraction pattern that ingests a
    document directly from a public HTTPS URL instead of uploading bytes. The
    workflow submits the document URL to the extraction queue, polls the shared
    jobs endpoint until the job is Processed, and fetches the inference to read
    the parsed fields. This is useful when documents already live in object
    storage or a CDN. 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: extract-from-url
  summary: Submit a document URL for extraction and read the parsed fields.
  description: >-
    Sends a publicly accessible document URL to the extraction enqueue
    endpoint, polls the job until processing finishes, and retrieves the
    extracted fields.
  inputs:
    type: object
    required:
    - authorization
    - modelId
    - url
    properties:
      authorization:
        type: string
        description: Mindee API key sent in the Authorization header.
      modelId:
        type: string
        description: UUID of the extraction model to apply.
      url:
        type: string
        description: Public HTTPS URL of the document to download and process.
      alias:
        type: string
        description: Optional free-form identifier tagged onto the request.
  steps:
  - stepId: enqueueByUrl
    description: >-
      Submit the document URL to the asynchronous extraction queue so Mindee
      downloads and processes the remote file.
    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
        url: $inputs.url
        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 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.enqueueByUrl.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 remote 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.enqueueByUrl.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      inferenceId: $response.body#/inference/id
      fields: $response.body#/inference/result/fields
  outputs:
    jobId: $steps.enqueueByUrl.outputs.jobId
    fields: $steps.getResult.outputs.fields