voicemate.audio.protocol

Control messages between the browser and the server, defined in one place.

Audio itself travels as binary WebSocket frames (see voicemate.audio.codec); everything else is a small JSON text message. static/audio.js implements the browser side, and a contract test keeps it in step with the definitions here.

Browser → server: MicOff, PlaybackStarted, PlaybackEnded. Server → browser: Stop.

Attributes

Classes

MicOff

The user turned the microphone off.

PlaybackStarted

The browser started playing assistant audio.

PlaybackEnded

The browser finished (or dropped) all queued assistant audio.

Stop

Barge-in: the browser must drop queued speech.

Functions

dump(message) → str

The JSON text sent over the WebSocket for message.

parse(text) → ControlMessage | None

Decode a JSON text message.

Module Contents

voicemate.audio.protocol.logger
class voicemate.audio.protocol.MicOff[source]

The user turned the microphone off.

type = 'mic'
state = 'off'
class voicemate.audio.protocol.PlaybackStarted[source]

The browser started playing assistant audio.

type = 'playback'
state = 'started'
class voicemate.audio.protocol.PlaybackEnded[source]

The browser finished (or dropped) all queued assistant audio.

type = 'playback'
state = 'ended'
class voicemate.audio.protocol.Stop[source]

Barge-in: the browser must drop queued speech.

type = 'stop'
state = None
voicemate.audio.protocol.ControlMessage
voicemate.audio.protocol.CONTROL_MESSAGES: tuple[ControlMessage, ...]
voicemate.audio.protocol.dump(message: ControlMessage) → str[source]

The JSON text sent over the WebSocket for message.

voicemate.audio.protocol.parse(text: str) → ControlMessage | None[source]

Decode a JSON text message.

Returns:

The message, or None (and a log line) when the text is malformed or unknown, so a bad client can never break the session.