Ashby Generate a Report and Poll for the Result

arazzo: 1.0.1
info:
  title: Ashby Generate a Report and Poll for the Result
  summary: Kick off an asynchronous report, poll it by request id until it completes, and take the data.
  description: >-
    Ashby exposes reporting two ways, and picking the wrong one is the classic
    mistake: report.synchronous returns data in the response but only survives
    small reports, while report.generate returns a requestId and an in_progress
    status that the caller must poll until it flips to complete or failed. This
    workflow drives the asynchronous path properly — it starts the generation,
    polls the same operation with the returned requestId, and branches on the
    terminal status rather than assuming success. 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: ashbyApi
  url: ../openapi/ashby-openapi.yml
  type: openapi
workflows:
- workflowId: report-generate-poll
  summary: Start an async report generation and poll by request id until it terminates.
  description: >-
    Starts a report generation, polls report.generate with the returned requestId
    until the status is complete or failed, and returns the report rows and
    column names on success.
  inputs:
    type: object
    required:
    - reportId
    properties:
      reportId:
        type: string
        description: The id of the report to generate.
      includeHeadersInData:
        type: boolean
        description: >-
          When true (the default), column headers are included as the first row
          of the data array. When false, headers appear only in columnNames.
  steps:
  - stepId: startReport
    description: >-
      Kick off the report generation. Ashby answers immediately with a requestId
      and a status, which is usually in_progress for any report large enough to
      justify this path over report.synchronous.
    operationId: reportGenerate
    requestBody:
      contentType: application/json
      payload:
        reportId: $inputs.reportId
        includeHeadersInData: $inputs.includeHeadersInData
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/success == true
    outputs:
      requestId: $response.body#/results/requestId
      initialStatus: $response.body#/results/status
    onSuccess:
    - name: alreadyComplete
      type: goto
      stepId: takeResult
      criteria:
      - context: $response.body
        condition: $.results.status == 'complete'
        type: jsonpath
    - name: stillRunning
      type: goto
      stepId: pollReport
      criteria:
      - context: $response.body
        condition: $.results.status == 'in_progress'
        type: jsonpath
  - stepId: pollReport
    description: >-
      Poll the same operation with the requestId from the start step. Supplying
      requestId is what turns report.generate from a kickoff into a status check;
      omitting it would queue a second report instead of polling the first. The
      retry backs off between attempts until the report reaches a terminal state.
    operationId: reportGenerate
    requestBody:
      contentType: application/json
      payload:
        reportId: $inputs.reportId
        requestId: $steps.startReport.outputs.requestId
        includeHeadersInData: $inputs.includeHeadersInData
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/success == true
    - context: $response.body
      condition: $.results.status != 'in_progress'
      type: jsonpath
    onSuccess:
    - name: reportComplete
      type: goto
      stepId: takeResult
      criteria:
      - context: $response.body
        condition: $.results.status == 'complete'
        type: jsonpath
    onFailure:
    - name: keepPolling
      type: retry
      retryAfter: 5
      retryLimit: 60
      criteria:
      - context: $response.body
        condition: $.results.status == 'in_progress'
        type: jsonpath
    outputs:
      status: $response.body#/results/status
      failureReason: $response.body#/results/failureReason
  - stepId: takeResult
    description: >-
      Take the finished report. This final call reads the completed generation by
      requestId and returns the rows and column names for the caller to load.
    operationId: reportGenerate
    requestBody:
      contentType: application/json
      payload:
        reportId: $inputs.reportId
        requestId: $steps.startReport.outputs.requestId
        includeHeadersInData: $inputs.includeHeadersInData
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/success == true
    - context: $response.body
      condition: $.results.status == 'complete'
      type: jsonpath
    outputs:
      status: $response.body#/results/status
      reportData: $response.body#/results/reportData
  outputs:
    requestId: $steps.startReport.outputs.requestId
    status: $steps.takeResult.outputs.status
    reportData: $steps.takeResult.outputs.reportData
    failureReason: $steps.pollReport.outputs.failureReason