arazzo: 1.0.1
info:
  title: Amazon Polly Start Synthesis Task and Poll to Completion
  summary: Start an async speech synthesis task, then poll it until it completes.
  description: >-
    The core asynchronous text-to-speech pattern in Amazon Polly. The workflow
    starts a SpeechSynthesisTask that writes its audio output to an S3 bucket,
    captures the returned task id, and then polls GetSpeechSynthesisTask in a
    loop until the task status leaves the scheduled or inProgress state. On
    completion it surfaces the S3 output URI. Each step spells out its request
    inline, including the AWS Signature Version 4 signing headers, so the flow
    can be read and executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: pollyApi
  url: ../openapi/amazon-polly-openapi-original.yaml
  type: openapi
workflows:
- workflowId: start-synthesis-task-poll
  summary: Start an asynchronous synthesis task and poll it until it reaches a terminal state.
  description: >-
    Submits a long-form synthesis task to an S3 bucket, captures the task id,
    and polls the task status until it is no longer scheduled or inProgress,
    branching back to poll again while still running.
  inputs:
    type: object
    required:
    - amzDate
    - authorization
    - text
    - outputFormat
    - outputS3BucketName
    - voiceId
    properties:
      amzDate:
        type: string
        description: The X-Amz-Date timestamp used to sign the requests.
      authorization:
        type: string
        description: The full SigV4 Authorization header value for the request.
      contentSha256:
        type: string
        description: The X-Amz-Content-Sha256 hex digest of the request payload.
      securityToken:
        type: string
        description: Optional X-Amz-Security-Token for temporary credentials.
      engine:
        type: string
        description: Engine to use for synthesis (standard or neural).
      text:
        type: string
        description: The input text (plain text or SSML) to synthesize.
      textType:
        type: string
        description: Whether the input text is plain text or ssml.
      outputFormat:
        type: string
        description: The audio output format (mp3, ogg_vorbis, pcm, or json).
      outputS3BucketName:
        type: string
        description: The S3 bucket name to which the output file will be saved.
      outputS3KeyPrefix:
        type: string
        description: Optional S3 key prefix for the output speech file.
      voiceId:
        type: string
        description: Voice ID to use for the synthesis (e.g. Joanna).
  steps:
  - stepId: startTask
    description: >-
      Start an asynchronous speech synthesis task that writes its output audio
      to the supplied S3 bucket, returning a task id to poll on.
    operationId: StartSpeechSynthesisTask
    parameters:
    - name: X-Amz-Date
      in: header
      value: $inputs.amzDate
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: X-Amz-Content-Sha256
      in: header
      value: $inputs.contentSha256
    - name: X-Amz-Security-Token
      in: header
      value: $inputs.securityToken
    requestBody:
      contentType: application/json
      payload:
        Engine: $inputs.engine
        OutputFormat: $inputs.outputFormat
        OutputS3BucketName: $inputs.outputS3BucketName
        OutputS3KeyPrefix: $inputs.outputS3KeyPrefix
        Text: $inputs.text
        TextType: $inputs.textType
        VoiceId: $inputs.voiceId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      taskId: $response.body#/SynthesisTask/TaskId
      taskStatus: $response.body#/SynthesisTask/TaskStatus
  - stepId: pollTask
    description: >-
      Poll the synthesis task by id, branching back to poll again while the task
      is still scheduled or in progress.
    operationId: GetSpeechSynthesisTask
    parameters:
    - name: TaskId
      in: path
      value: $steps.startTask.outputs.taskId
    - name: X-Amz-Date
      in: header
      value: $inputs.amzDate
    - name: Authorization
      in: header
      value: $inputs.authorization
    - name: X-Amz-Content-Sha256
      in: header
      value: $inputs.contentSha256
    - name: X-Amz-Security-Token
      in: header
      value: $inputs.securityToken
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      taskStatus: $response.body#/SynthesisTask/TaskStatus
      outputUri: $response.body#/SynthesisTask/OutputUri
      taskStatusReason: $response.body#/SynthesisTask/TaskStatusReason
    onSuccess:
    - name: stillRunning
      type: goto
      stepId: pollTask
      criteria:
      - context: $response.body
        condition: $.SynthesisTask.TaskStatus == "scheduled" || $.SynthesisTask.TaskStatus == "inProgress"
        type: jsonpath
    - name: terminal
      type: end
      criteria:
      - context: $response.body
        condition: $.SynthesisTask.TaskStatus == "completed" || $.SynthesisTask.TaskStatus == "failed"
        type: jsonpath
  outputs:
    taskId: $steps.startTask.outputs.taskId
    taskStatus: $steps.pollTask.outputs.taskStatus
    outputUri: $steps.pollTask.outputs.outputUri