arazzo: 1.0.1
info:
  title: Amazon Forecast Group Then Train
  summary: Create a dataset group, wait until ACTIVE, then train a predictor on it.
  description: >-
    Bridges dataset-group provisioning into model training. The workflow creates
    a dataset group from existing dataset ARNs, polls the dataset group listing
    until the group reports an ACTIVE status, and then creates a predictor whose
    input data configuration trains against that group. Because the Forecast
    description only exposes a describe operation for datasets, dataset group
    status is read from listDatasetGroups. 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: amazonForecastApi
  url: ../openapi/amazon-forecast-openapi.yml
  type: openapi
workflows:
- workflowId: group-then-train
  summary: Create a dataset group, wait until ACTIVE, then train a predictor.
  description: >-
    Creates a dataset group from supplied dataset ARNs, polls listDatasetGroups
    until the group is ACTIVE, and then creates a predictor trained on it.
  inputs:
    type: object
    required:
    - datasetGroupName
    - domain
    - datasetArns
    - predictorName
    - forecastHorizon
    - inputDataConfig
    - featurizationConfig
    properties:
      datasetGroupName:
        type: string
        description: A name for the dataset group.
      domain:
        type: string
        description: The domain associated with the dataset group.
      datasetArns:
        type: array
        description: The dataset ARNs to include in the group.
        items:
          type: string
      predictorName:
        type: string
        description: A name for the predictor.
      forecastHorizon:
        type: integer
        description: The number of time-steps to predict.
      inputDataConfig:
        type: object
        description: The training input configuration referencing the dataset group.
      featurizationConfig:
        type: object
        description: The featurization configuration for the predictor.
  steps:
  - stepId: createDatasetGroup
    description: >-
      Create the dataset group from the supplied dataset ARNs.
    operationId: createDatasetGroup
    requestBody:
      contentType: application/json
      payload:
        DatasetGroupName: $inputs.datasetGroupName
        Domain: $inputs.domain
        DatasetArns: $inputs.datasetArns
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      datasetGroupArn: $response.body#/DatasetGroupArn
  - stepId: waitForGroupActive
    description: >-
      List dataset groups and confirm the newly created group has reached the
      ACTIVE status. Repeat this step while the group is still being created.
    operationId: listDatasetGroups
    successCriteria:
    - condition: $statusCode == 200
    - context: $response.body
      condition: $.DatasetGroups[?(@.DatasetGroupArn == "$steps.createDatasetGroup.outputs.datasetGroupArn")].Status contains "ACTIVE"
      type: jsonpath
    onSuccess:
    - name: groupActive
      type: goto
      stepId: createPredictor
      criteria:
      - context: $response.body
        condition: $.DatasetGroups[?(@.DatasetGroupArn == "$steps.createDatasetGroup.outputs.datasetGroupArn")].Status contains "ACTIVE"
        type: jsonpath
    onFailure:
    - name: stillCreating
      type: retry
      retryAfter: 30
      retryLimit: 20
      criteria:
      - condition: $statusCode == 200
    outputs:
      datasetGroups: $response.body#/DatasetGroups
  - stepId: createPredictor
    description: >-
      Train a predictor against the now-ACTIVE dataset group.
    operationId: createPredictor
    requestBody:
      contentType: application/json
      payload:
        PredictorName: $inputs.predictorName
        ForecastHorizon: $inputs.forecastHorizon
        InputDataConfig: $inputs.inputDataConfig
        FeaturizationConfig: $inputs.featurizationConfig
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      predictorArn: $response.body#/PredictorArn
  outputs:
    datasetGroupArn: $steps.createDatasetGroup.outputs.datasetGroupArn
    predictorArn: $steps.createPredictor.outputs.predictorArn