Skip to content

VoxtaAudioClient

The VoxtaAudioClient is a specialized client designed for high-throughput, low-latency binary audio streaming. While the main VoxtaClient handles text and control messages, the VoxtaAudioClient establishes a dedicated WebSocket for raw PCM data.

Typical Use Case

You would use this client if you are building a voice-to-voice application where you need to stream the user's microphone to Voxta (STT) and receive the character's voice as a stream (TTS).

Dedicated client for handling binary PCM audio streaming to/from Voxta.

Initialize the audio client.

Parameters:

  • url (str) –

    The base URL of the Voxta server.

  • logger (Optional[Logger], default: None ) –

    Optional logger instance.

Attributes

running instance-attribute 

running = False

Functions

connect async 

connect(connection_token: str, cookies: Optional[dict[str, str]] = None)

Connect to the audio stream WebSocket and authenticate.

Parameters:

  • connection_token (str) –

    The SignalR connection token obtained from negotiation.

  • cookies (Optional[dict[str, str]], default: None ) –

    Optional cookies to include in the connection request.

send_audio async 

send_audio(pcm_data: bytes)

Send binary PCM data to the server.

Parameters:

  • pcm_data (bytes) –

    Binary PCM audio data.

on_audio 

on_audio(callback: Callable[[bytes], None])

Register a callback for received audio data.

Parameters:

  • callback (Callable[[bytes], None]) –

    Function that receives binary PCM data.

close async 

close()

Close the audio client connection.

Audio Format

Currently, the audio client expects and delivers:

  • Format: Raw PCM
  • Sample Rate: Depends on server configuration (typically 16kHz or 24kHz)
  • Channels: Mono
  • Bit Depth: 16-bit signed integer

Implementation Example

audio_client = VoxtaAudioClient("http://localhost:5384")

# Handle incoming audio from the AI
@audio_client.on_audio
def handle_tts(pcm_data):
    # Play the data via sounddevice or save to a buffer
    pass

# Start connection (requires token from VoxtaClient.negotiate)
await audio_client.connect(token)

# Send user voice data
await audio_client.send_audio(microphone_bytes)