arazzo: 1.0.1
info:
  title: Amazon Polly Store Lexicon, Start Async Task, and Poll
  summary: Store a lexicon, start an async synthesis task applying it, then poll to completion.
  description: >-
    Combines custom pronunciation with long-form asynchronous synthesis. The
    workflow stores a PLS lexicon with PutLexicon, starts a SpeechSynthesisTask
    that applies that lexicon and writes audio to an S3 bucket, and then polls
    GetSpeechSynthesisTask until the task reaches a terminal state, surfacing 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: lexicon-async-task-poll
  summary: Store a lexicon, start a synthesis task that applies it, and poll to completion.
  description: >-
    Stores a PLS lexicon, starts an asynchronous synthesis task that applies the
    lexicon and writes to S3, and polls the task status until it completes or
    fails.
  inputs:
    type: object
    required:
    - lexiconName
    - content
    - amzDate
    - authorization
    - text
    - outputFormat
    - outputS3BucketName
    - voiceId
    properties:
      lexiconName:
        type: string
        description: Case-sensitive alphanumeric lexicon name (1-20 chars).
      content:
        type: string
        description: The PLS lexicon content as a string.
      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.
      text:
        type: string
        description: The input text (plain text or SSML) to synthesize.
      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.
      voiceId:
        type: string
        description: Voice ID to use for the synthesis (e.g. Joanna).
  steps:
  - stepId: putLexicon
    description: Store the PLS lexicon to be applied during the synthesis task.
    operationId: PutLexicon
    parameters:
    - name: LexiconName
      in: path
      value: $inputs.lexiconName
    - 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:
        Content: $inputs.content
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $statusCode
  - stepId: startTask
    description: >-
      Start an asynchronous synthesis task that applies the stored lexicon and
      writes its output to the supplied S3 bucket.
    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:
        LexiconNames:
        - $inputs.lexiconName
        OutputFormat: $inputs.outputFormat
        OutputS3BucketName: $inputs.outputS3BucketName
        Text: $inputs.text
        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 it 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
    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