My code can make an exchange like :

1 question (by the user) –> 1 answer by my llm.

How can I make the conversation continue ?

1 question (by the user) –> 1 answer by my llm.-> 1 question (by the user) -> 1 answer by the llm etc..

@sock.route(WEBSOCKET_ROUTE)
def transcription_websocket(ws):
    exchange_count = 0
    max_exchanges = 5  
    has_seen_media =  False 
    while exchange_count < max_exchanges:
        data = json.loads(ws.receive())
        match data['event']:
            case "connected":
                transcriber = TwilioTranscriber()
                transcriber.connect()
                print('transcriber connected')
            case "start":
                print('twilio started')
                has_seen_media =  False 
            case "media": 
                if not has_seen_media:
                    payload_b64 = data['media']['payload']
                    payload_mulaw = base64.b64decode(payload_b64)
                    transcriber.stream(payload_mulaw)
                    twilio = transcriber.llm
                    ngrok = listener.url()
                    if transcriber.llm_call_in_progress and twilio : 
                        send_tts_message(current_calls["call_sid"],twilio,ngrok)
                        twilio = None 
                        transcriber.llm_call_in_progress = False
                        exchange_count += 1 
                        has_seen_media = True 
                        if exchange_count >= max_exchanges:
                            ws.send(json.dumps({"continue": True}))  
                            exchange_count = 0 


def send_tts_message(call_sid, message, ngrok):
    client = Client(api_key, api_secret, account_sid)
    # TwiML with a Gather that sends a callback to your server endpoint
    twiml = f'<Response><Say>{message}</Say><Gather numDigits="0" timeout="10"><Pause length="10"/></Gather></Response>'
    call = client.calls(call_sid).update(twiml=twiml)
    print(f"Message sent to call {call_sid}")

The transcription_websocket handles the transcription (of what is said through the phone) and the response of the llm.

We tried to change the twML using a Gather (Without result) we also tried a “for” loop, that didn’t work.

We would be really grateful of any response !!