String/Bytes Type Mismatch in WebSocket Communication

[dnhrdt]

There is a type mismatch between the server's expected data type and the client's sent data type when handling the end-of-audio signal.

Issue

The server's get_audio_from_websocket() method expects binary data and checks for b"END_OF_AUDIO":

def get_audio_from_websocket(self, websocket):
    frame_data = websocket.recv()
    if frame_data == b"END_OF_AUDIO":  # Expects bytes
        return False
    return np.frombuffer(frame_data, dtype=np.float32)

However, the ServeClientBase.disconnect() method sends a string message:

def disconnect(self):
    self.websocket.send(json.dumps({  # Sends string
        "uid": self.client_uid,
        "message": self.DISCONNECT
    }))

This causes a "bytes-like object is required, not 'str'" error when the server tries to process the disconnect message.

Steps to Reproduce

1. Start WhisperLive server
2. Connect client
3. Send audio data
4. Try to disconnect client
5. Observe error in server logs: "Unexpected error: a bytes-like object is required, not 'str'"

Expected Behavior

The server should either:

1. Accept string messages for control signals like disconnect
2. Or document that all WebSocket communication must be binary data

Suggested Fix

Either:

Modify disconnect() to send binary data:

def disconnect(self):
    self.websocket.send(b"END_OF_AUDIO")

Or modify get_audio_from_websocket() to handle both binary and string data:

def get_audio_from_websocket(self, websocket):
    frame_data = websocket.recv()
    if isinstance(frame_data, str):
        try:
            control_msg = json.loads(frame_data)
            if control_msg.get("message") == self.DISCONNECT:
                return False
        except json.JSONDecodeError:
            pass
    elif frame_data == b"END_OF_AUDIO":
        return False
    return np.frombuffer(frame_data, dtype=np.float32)

[makaveli10]

@dnhrdt Thanks for raising the issue. Feel free to open a PR with the fix.

[makaveli10]

@dnhrdt I dont see this error "Unexpected error: a bytes-like object is required, not 'str'" on my end if I disconnect the client with a KeyboardInterrupt. Can you please be more specific on how to reproduce this?

What I see on the server side if do a KeyboardInterrupt on the client side:

INFO:faster_whisper:Processing audio with duration 00:02.816
INFO:faster_whisper:VAD filter removed 00:00.000 of audio
INFO:root:Connection closed by client
INFO:root:Cleaning up.
ERROR:root:[ERROR]: Sending data to client: received 1000 (OK); then sent 1000 (OK)
INFO:root:Exiting speech to text thread