Deepgram Transcribe Audio and Track Usage

arazzo: 1.0.1
info:
  title: Deepgram Transcribe Audio and Track Usage
  summary: Transcribe a pre-recorded audio URL and then reconcile the request against project usage and request logs.
  description: >-
    A common observability pattern for speech-to-text pipelines. The workflow
    submits a pre-recorded audio URL for transcription, captures the returned
    request_id, and then queries the project's usage summary and request log so
    the freshly submitted transcription can be reconciled against billing and
    audit data. 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: speechToTextApi
  url: ../openapi/deepgram-speech-to-text-openapi.yml
  type: openapi
- name: managementApi
  url: ../openapi/deepgram-management-openapi.yml
  type: openapi
workflows:
- workflowId: transcribe-and-track-usage
  summary: Transcribe an audio URL then reconcile it against project usage and requests.
  description: >-
    Transcribes a hosted audio file, then retrieves the project usage summary
    and the recent request log for the time window so the transcription can be
    audited against billing data.
  inputs:
    type: object
    required:
    - apiKey
    - projectId
    - audioUrl
    properties:
      apiKey:
        type: string
        description: Deepgram API key used to authenticate requests.
      projectId:
        type: string
        description: The Deepgram project the usage and request logs belong to.
      audioUrl:
        type: string
        description: Publicly accessible URL of the audio file to transcribe.
      model:
        type: string
        description: Speech-to-text model to use for transcription (e.g. nova-3).
        default: nova-3
      start:
        type: string
        description: Start of the usage/request window in ISO 8601 format.
      end:
        type: string
        description: End of the usage/request window in ISO 8601 format.
  steps:
  - stepId: transcribeAudio
    description: >-
      Submit the hosted audio URL for transcription with punctuation and smart
      formatting enabled, capturing the request identifier from the metadata.
    operationId: transcribePreRecordedAudio
    parameters:
    - name: Authorization
      in: header
      value: Token $inputs.apiKey
    - name: model
      in: query
      value: $inputs.model
    - name: punctuate
      in: query
      value: true
    - name: smart_format
      in: query
      value: true
    requestBody:
      contentType: application/json
      payload:
        url: $inputs.audioUrl
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      requestId: $response.body#/metadata/request_id
      duration: $response.body#/metadata/duration
      transcript: $response.body#/results/channels/0/alternatives/0/transcript
  - stepId: getUsage
    description: >-
      Retrieve the aggregated usage summary for the project over the supplied
      time window to confirm the audio hours were recorded.
    operationId: getProjectUsage
    parameters:
    - name: Authorization
      in: header
      value: Token $inputs.apiKey
    - name: project_id
      in: path
      value: $inputs.projectId
    - name: start
      in: query
      value: $inputs.start
    - name: end
      in: query
      value: $inputs.end
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      usageStart: $response.body#/start
      usageEnd: $response.body#/end
      usageResults: $response.body#/results
  - stepId: listRequests
    description: >-
      List the project's recent API requests for the same window so the
      transcription request_id can be located in the audit log.
    operationId: listProjectRequests
    parameters:
    - name: Authorization
      in: header
      value: Token $inputs.apiKey
    - name: project_id
      in: path
      value: $inputs.projectId
    - name: start
      in: query
      value: $inputs.start
    - name: end
      in: query
      value: $inputs.end
    - name: limit
      in: query
      value: 50
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      requests: $response.body#/requests
      firstRequestId: $response.body#/requests/0/request_id
  outputs:
    requestId: $steps.transcribeAudio.outputs.requestId
    transcript: $steps.transcribeAudio.outputs.transcript
    usageResults: $steps.getUsage.outputs.usageResults
    requests: $steps.listRequests.outputs.requests