arazzo: 1.0.1
info:
  title: Databricks Trigger Job Run and Wait for Output
  summary: Trigger a job run, poll the run until terminal, then fetch its output.
  description: >-
    Triggers an immediate run of an existing Databricks job, polls the run
    metadata until the run reaches a TERMINATED life cycle state, and then
    retrieves the run output (including the notebook exit value or error trace).
    The run_id returned by run-now drives both the poll loop and the final
    output fetch. 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: databricksApi
  url: ../openapi/databricks-openapi.yml
  type: openapi
workflows:
- workflowId: run-job-and-wait
  summary: Run a job now, wait for completion, and return its output.
  description: >-
    Submits a run-now request for a job, repeatedly reads the run until its
    life cycle state is TERMINATED, then pulls the run output for the result
    or error details.
  inputs:
    type: object
    required:
    - job_id
    properties:
      job_id:
        type: integer
        description: The canonical identifier of the job to run.
      notebook_params:
        type: object
        description: Optional map of notebook parameter name/value pairs.
      idempotency_token:
        type: string
        description: Optional token guaranteeing run idempotency.
  steps:
  - stepId: runJobNow
    description: >-
      Trigger an immediate run of the job, optionally overriding notebook
      parameters. Returns the globally unique run_id.
    operationId: runJobNow
    requestBody:
      contentType: application/json
      payload:
        job_id: $inputs.job_id
        notebook_params: $inputs.notebook_params
        idempotency_token: $inputs.idempotency_token
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      runId: $response.body#/run_id
      numberInJob: $response.body#/number_in_job
  - stepId: pollRun
    description: >-
      Read the run metadata and inspect the life cycle state. Loop back while
      the run is PENDING or RUNNING; continue once it is TERMINATED.
    operationId: getJobRun
    parameters:
    - name: run_id
      in: query
      value: $steps.runJobNow.outputs.runId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      lifeCycleState: $response.body#/state/life_cycle_state
      resultState: $response.body#/state/result_state
      runPageUrl: $response.body#/run_page_url
    onSuccess:
    - name: stillRunning
      type: goto
      stepId: pollRun
      criteria:
      - context: $response.body
        condition: $.state.life_cycle_state == "RUNNING"
        type: jsonpath
    - name: stillPending
      type: goto
      stepId: pollRun
      criteria:
      - context: $response.body
        condition: $.state.life_cycle_state == "PENDING"
        type: jsonpath
    - name: terminated
      type: goto
      stepId: getRunOutput
      criteria:
      - context: $response.body
        condition: $.state.life_cycle_state == "TERMINATED"
        type: jsonpath
  - stepId: getRunOutput
    description: >-
      Retrieve the output and metadata of the completed run, exposing the
      notebook result and any error trace.
    operationId: getJobRunOutput
    parameters:
    - name: run_id
      in: query
      value: $steps.runJobNow.outputs.runId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      notebookResult: $response.body#/notebook_output/result
      error: $response.body#/error
  outputs:
    runId: $steps.runJobNow.outputs.runId
    resultState: $steps.pollRun.outputs.resultState
    notebookResult: $steps.getRunOutput.outputs.notebookResult