speech-to-text/extension.yaml

name: speech-to-text
version: 0.1.6
specVersion: v1beta

displayName: Transcribe Speech to Text
description:
  Transcribes audio files in Cloud Storage to .txt files using Cloud Speech To
  Text.

icon: icon.png

tags: [ai, speech-to-text, speech-transcription, google-ai]

license: Apache-2.0 # The license you want for the extension

author:
  authorName: Google Cloud
  url: https://cloud.google.com/

sourceUrl: https://github.com/GoogleCloudPlatform/firebase-extensions/tree/main/speech-to-text
releaseNotesUrl: https://github.com/GoogleCloudPlatform/firebase-extensions/tree/main/speech-to-text/CHANGELOG.md

apis:
  - apiName: speech.googleapis.com
    reason: Used for transcribing the audio of sound files.

resources:
  - name: transcribeAudio
    type: firebaseextensions.v1beta.function
    description: >-
      Listens for new audio files uploaded to a specified Cloud Storage bucket,
      transcribes the speech in those files, then stores the transcription in
      storage, or in firestore, or in both.
    properties:
      location: ${param:LOCATION}
      runtime: nodejs20
      availableMemoryMb: 1024
      eventTrigger:
        eventType: google.storage.object.finalize
        resource: projects/_/buckets/${param:EXTENSION_BUCKET}

params:
  - param: LOCATION
    label: Cloud Functions location
    description: >-
      Where do you want to deploy the functions created for this extension? You
      usually want a location close to your database. Realtime Database
      instances are located in `us-central1`. For help selecting a location,
      refer to the [location selection
      guide](https://firebase.google.com/docs/functions/locations).
    type: select
    options:
      - label: Iowa (us-central1)
        value: us-central1
      - label: South Carolina (us-east1)
        value: us-east1
      - label: Northern Virginia (us-east4)
        value: us-east4
      - label: Belgium (europe-west1)
        value: europe-west1
      - label: London (europe-west2)
        value: europe-west2
      - label: Frankfurt (europe-west3)
        value: europe-west3
      - label: Hong Kong (asia-east2)
        value: asia-east2
      - label: Tokyo (asia-northeast1)
        value: asia-northeast1
    default: us-central1
    required: true
    immutable: true

  - param: EXTENSION_BUCKET
    label: Cloud Storage bucket for input and output
    description: >
      The Cloud Storage bucket that the extension should be listening to. Files
      uploaded to this bucket will be transcribed by the extension. If cloud
      storage output is enabled, transcriptions will be written to this bucket.
    type: selectResource
    resourceType: storage.googleapis.com/Bucket
    example: my-project-12345.appspot.com
    validationRegex: ^([0-9a-z_.-]*)$
    validationErrorMessage: Invalid storage bucket
    default: ${STORAGE_BUCKET}
    required: true

  - param: OUTPUT_STORAGE_PATH
    label: Storage path for transcriptions
    description: >
      The storage path in which to output transcriptions. If this is not set,
      the extension will output to the root of the bucket.
    example: transcriptions
    required: false

  - param: COLLECTION_PATH
    label: Firestore collection for storing transcribed audio
    description: >
      The firestore collection in which to output transcriptions. If this is not
      set, the extension will not output the data to Firestore.
    validationRegex: '^[^/]+(/[^/]+/[^/]+)*$'
    validationErrorMessage: Must be a valid Cloud Firestore Collection
    example: transcriptions
    required: false

  - param: LANGUAGE_CODE
    label: BCP-47 code of the transcription language
    description: >
      The BCP-47 code of the transcription language, as shown in the [Language
      support
      documentation](https://cloud.google.com/speech-to-text/docs/languages)
    example: e.g. en-US
    # TODO(reao): Here I assume that language codes need to end in two
    #   uppercase characters -- is that correct? It is a pattern in the language
    #   support documentation.
    validationRegex: '^([a-zA-Z-])*[A-Z][A-Z]$'
    validationErrorMessage:
      Must be a valid code from
      https://cloud.google.com/speech-to-text/docs/languages
    required: true

  - param: MODEL
    label: Language model used for transcription
    description: >
      Which kind of use-case should the speech-to-text transcription algorithm
      be honed for? For details, see [the model field in the
      documentation](https://cloud.google.com/speech-to-text/docs/reference/rest/v1/RecognitionConfig)

      If you're not sure, just use the default.
    example: default
    default: default
    required: true

  - param: ENABLE_AUTOMATIC_PUNCTUATION
    label: Enable automatic punctuation
    description: >
      Should the transcription algorithm attempt to add punctuation to the
      transcription? For details, see [the
      documentation](https://cloud.google.com/speech-to-text/docs/automatic-punctuation)
    type: select
    options:
      - label: Enabled
        value: true
      - label: Disabled
        value: false
    default: true

roles:
  - role: storage.objectAdmin
    reason: Allows the extension to write to your Cloud Storage.
  - role: datastore.user
    reason: Allows the extension to write to your Firestore Database instance.

events:
  - type: firebase.extensions.storage-transcribe-audio.v1.complete
    description:
      Occurs when audio transcription completes. The event will contain the full
      transcription.
  - type: firebase.extensions.storage-transcribe-audio.v1.fail
    description:
      Occurs when audio transcription fails. The event will contain information
      about the failure.