Conversation Bundle Module

The viam:conversation-bundle module provides these models for conversational workflows:

1. viam:conversation-bundle:text-to-speech - A generic service that synthesises speech via Google Cloud Text-to-Speech and plays it through an audio_out component.

---

Model: viam:conversation-bundle:text-to-speech

API: rdk:service:generic

Synthesises speech using the Google Cloud Text-to-Speech API and plays the resulting audio through an rdk:component:audio_out component.

Prerequisites

• A Google Cloud project with the Text-to-Speech API enabled.
• A service account key (JSON) with access to the API.
• A configured audio_out component on the same machine.

Configuration

{
  "audio_out": "<string>",
  "google_credentials_json": { ... },
  "language_code": "<string>",
  "voice_name": "<string>"
}

Name	Type	Required	Description
audio_out	string	Yes	Name of the audio_out component dependency used for playback.
google_credentials_json	object	Yes	Google Cloud service account credentials as a JSON object (not a string).
language_code	string	No	BCP-47 language code. Defaults to "en-US".
voice_name	string	No	Specific Google voice name (e.g. "en-US-Neural2-F"). If omitted, Google picks a default for the language.

Example Configuration

{
  "audio_out": "ao",
  "google_credentials_json": {
    "type": "service_account",
    "project_id": "my-project",
    "private_key_id": "abc123",
    "private_key": "-----BEGIN PRIVATE KEY-----\n...\n-----END PRIVATE KEY-----\n",
    "client_email": "[email protected]",
    "client_id": "123456789",
    "auth_uri": "https://accounts.google.com/o/oauth2/auth",
    "token_uri": "https://oauth2.googleapis.com/token"
  },
  "language_code": "en-US",
  "voice_name": "en-US-Neural2-F"
}

DoCommand

say — Synthesise and play text. The call blocks until playback completes.

{"say": "Hello, your espresso is ready!"}

Returns:

{"text": "Hello, your espresso is ready!"}

say_async — Queue text for playback and return immediately without waiting for synthesis or playback to finish. A background worker drains the queue and plays items sequentially. Audio is only sent to the speaker when no other speech (sync or async) is currently playing, so queued messages will never overlap with an in-flight say call. Returns an error if the async queue is full (capacity 64).

{"say_async": "Hello, your espresso is ready!"}

Returns:

{"queued": "Hello, your espresso is ready!"}