arazzo: 1.0.1
info:
  title: Amazon Neptune ML Model Transform Job
  summary: Launch a Neptune ML model transform job from a trained model and poll it to completion.
  description: >-
    Generates model artifacts for inference from a completed model training job.
    The workflow creates a model transform job that reads a training job's output,
    polls the transform job status on a loop until it is Completed, and then lists
    the active transform jobs to confirm the job ran. The poll loop uses a retry
    delay to handle the in-progress state. 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: neptuneMlApi
  url: ../openapi/amazon-neptune-ml-openapi.yml
  type: openapi
workflows:
- workflowId: ml-model-transform
  summary: Create a model transform job, poll until Completed, and confirm it is listed.
  description: >-
    Creates a model transform job from a training job, polls until it completes,
    and lists transform jobs.
  inputs:
    type: object
    required:
    - mlModelTrainingJobId
    - modelTransformOutputS3Location
    properties:
      mlModelTrainingJobId:
        type: string
        description: Job ID of the completed model training job.
      modelTransformOutputS3Location:
        type: string
        description: S3 location for transform output artifacts.
  steps:
  - stepId: startTransform
    description: >-
      Create a model transform job that reads the training job's output and
      writes transform artifacts to S3.
    operationId: createModelTransformJob
    requestBody:
      contentType: application/json
      payload:
        mlModelTrainingJobId: $inputs.mlModelTrainingJobId
        modelTransformOutputS3Location: $inputs.modelTransformOutputS3Location
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      transformJobId: $response.body#/id
  - stepId: pollTransform
    description: >-
      Poll the transform job status. Retry while it is still in progress and
      continue once its status is Completed.
    operationId: getModelTransformJobStatus
    parameters:
    - name: id
      in: path
      value: $steps.startTransform.outputs.transformJobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      transformStatus: $response.body#/status
    onSuccess:
    - name: transformRunning
      type: retry
      retryAfter: 60
      retryLimit: 120
      criteria:
      - context: $response.body
        condition: $.status != "Completed"
        type: jsonpath
    - name: transformDone
      type: goto
      stepId: listTransforms
      criteria:
      - context: $response.body
        condition: $.status == "Completed"
        type: jsonpath
  - stepId: listTransforms
    description: >-
      List the active model transform jobs to confirm the transform job ran.
    operationId: listModelTransformJobs
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      transformJobIds: $response.body#/ids
  outputs:
    transformJobId: $steps.startTransform.outputs.transformJobId
    transformStatus: $steps.pollTransform.outputs.transformStatus