arazzo: 1.0.1
info:
  title: Amazon Fraud Detector Provision Model and Detector
  summary: Define an event type, create an ML model and a detector on top of it, then confirm the detector exists.
  description: >-
    Builds the core fraud detection assets in order: it defines (or updates) the
    event type that scopes the schema, creates a machine learning model bound to
    that event type, creates a detector that orchestrates models and rules for
    the same event type, and finally reads the detector back to confirm it was
    created. Each 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: fraudDetectorApi
  url: ../openapi/amazon-fraud-detector-openapi.yml
  type: openapi
workflows:
- workflowId: provision-model-detector
  summary: Provision an event type, an ML model, and a detector in one pass.
  description: >-
    Defines the event type, creates an ML model and a detector that both
    reference it, then reads the detector back to verify it exists.
  inputs:
    type: object
    required:
    - eventTypeName
    - eventVariables
    - labels
    - entityTypes
    - modelId
    - modelType
    - detectorId
    properties:
      eventTypeName:
        type: string
        description: The event type that the model and detector both operate on.
      eventVariables:
        type: array
        description: The event variable names that make up the event schema.
        items:
          type: string
      labels:
        type: array
        description: The label names available for this event type.
        items:
          type: string
      entityTypes:
        type: array
        description: The entity type names associated with the event type.
        items:
          type: string
      modelId:
        type: string
        description: The identifier for the ML model to create (e.g. "fraud-model-v1").
      modelType:
        type: string
        description: The model type, such as ONLINE_FRAUD_INSIGHTS.
      modelDescription:
        type: string
        description: A human readable description of the model.
      detectorId:
        type: string
        description: The identifier for the detector to create (e.g. "payment-fraud-detector").
      detectorDescription:
        type: string
        description: A human readable description of the detector.
  steps:
  - stepId: defineEventType
    description: >-
      Create or update the event type so the model and detector have a schema to
      bind to.
    operationId: putEventType
    requestBody:
      contentType: application/json
      payload:
        name: $inputs.eventTypeName
        eventVariables: $inputs.eventVariables
        labels: $inputs.labels
        entityTypes: $inputs.entityTypes
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      eventTypeName: $inputs.eventTypeName
  - stepId: createModel
    description: >-
      Create the ML model bound to the event type defined in the previous step.
    operationId: putModel
    requestBody:
      contentType: application/json
      payload:
        modelId: $inputs.modelId
        modelType: $inputs.modelType
        description: $inputs.modelDescription
        eventTypeName: $steps.defineEventType.outputs.eventTypeName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      modelId: $inputs.modelId
  - stepId: createDetector
    description: >-
      Create the detector that orchestrates models and rules for the same event
      type.
    operationId: putDetector
    requestBody:
      contentType: application/json
      payload:
        detectorId: $inputs.detectorId
        description: $inputs.detectorDescription
        eventTypeName: $steps.defineEventType.outputs.eventTypeName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      detectorId: $inputs.detectorId
  - stepId: confirmDetector
    description: >-
      Read the detector back by id to confirm it was created against the
      expected event type.
    operationId: getDetectors
    requestBody:
      contentType: application/json
      payload:
        detectorId: $inputs.detectorId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      detectorId: $response.body#/detectors/0/detectorId
      detectorArn: $response.body#/detectors/0/arn
  outputs:
    modelId: $steps.createModel.outputs.modelId
    detectorId: $steps.confirmDetector.outputs.detectorId
    detectorArn: $steps.confirmDetector.outputs.detectorArn