[Bug]: Rev AI transcriber not working, abruptly disconnecting

[raghavjaistra]

Brief Description

When I use Azure as a transcriber, things are working perfectly, but as soon as I change it to RevAI, it is not able to recognise the audio packets and is disconnected in between (got to know after extensive logging)

telephony_server = TelephonyServer(
    base_url=BASE_URL,
    config_manager=config_manager,
    inbound_call_configs=[
        TwilioInboundCallConfig(
            url="/inbound_call",
            agent_config=ChatGPTAgentConfig(
                openai_api_key=OPENAI_API_KEY,
                model_name="gpt-4o",
                initial_message=BaseMessage(text="Hi, I am your customer support agent. How may i help you today?"),
                prompt_preamble="""You are a customer support agent. Your role is to provide resolution to a customer that is on call with you. You can make-up information to play along, to do this conversation. Just make it sound like a real conversation on telephone. Important: Generate short responses.""",
            ),
            twilio_config=TwilioConfig(
                account_sid=os.environ["TWILIO_ACCOUNT_SID"],
                auth_token=os.environ["TWILIO_AUTH_TOKEN"],
            ),
            synthesizer_config=AzureSynthesizerConfig(
                voice_name="en-GB-LibbyNeural",
                sampling_rate=DEFAULT_SAMPLING_RATE,
                audio_encoding=DEFAULT_AUDIO_ENCODING,
            ),
            transcriber_config=RevAITranscriberConfig.from_telephone_input_device(
            ),
        )
    ],
)

Although this plugged in, is working perfectly.

transcriber_config=AzureTranscriberConfig.from_telephone_input_device(
                candidate_languages=["en-US"],
            )

The REV_AI_API_KEY is also set and this is the log I am getting after logging into threv_ai_transcriber.

Rev AI process called
main-telephony-app-1  | Rev AI url:  wss://api.rev.ai/speechtotext/v1/stream?access_token=02Em5IbtqKEgiKZOIN1O7z_hRZ5-M2JPBdMW86tdN1nccJ1ujSFK4ps2JfMuU3-pp9j9rqMq2HYRECvXQsoyHf9PfvMiA&content_type=audio/x-raw;layout=interleaved;rate=16000;format=S16LE;channels=1
main-telephony-app-1  | Processing agent response
main-telephony-app-1  | 1733389100.974085 Processing agent response: message=EndOfTurn() is_interruptible=True is_first=False is_sole_text_chunk=False
main-telephony-app-1  | Sending end of turn
main-telephony-app-1  | Creating interruptible agent response event
main-telephony-app-1  | Processing agent response
main-telephony-app-1  | 1733389100.974721 Processing agent response: message=BaseMessage(text='Traffic Violations Helpline. Please wait while i authenticate you..', trailing_silence_seconds=0.0, cache_phrase=None) is_interruptible=True is_first=False is_sole_text_chunk=False
main-telephony-app-1  | 1733389100.974768 Synthesizing speech for message
main-telephony-app-1  | Creating interruptible agent response event
main-telephony-app-1  | Start sending speech Traffic Violations Helpline. Please wait while i authenticate you.. to output
main-telephony-app-1  | INFO:     10.131.45.73:49124 - "GET / HTTP/1.1" 404 Not Found
main-telephony-app-1  | Traceback (most recent call last):
main-telephony-app-1  |   File "/usr/local/lib/python3.10/site-packages/vocode/streaming/transcriber/rev_ai_transcriber.py", line 102, in receiver
main-telephony-app-1  |     msg = await ws.recv()
main-telephony-app-1  |   File "/usr/local/lib/python3.10/site-packages/websockets/legacy/protocol.py", line 568, in recv
main-telephony-app-1  |     await self.ensure_open()
main-telephony-app-1  |   File "/usr/local/lib/python3.10/site-packages/websockets/legacy/protocol.py", line 948, in ensure_open
main-telephony-app-1  |     raise self.connection_closed_exc()
main-telephony-app-1  | websockets.exceptions.ConnectionClosedOK: sent 1000 (OK); then received 1000 (OK) Closing connection
main-telephony-app-1  | Received message from Rev AI: {"type":"connected","id":"0g9J1oA4qYmdbqDv"}
main-telephony-app-1  | Received 'connected' message from Rev AI.
main-telephony-app-1  | 1733389101.548895 Finished sending all chunks to Twilio
main-telephony-app-1  | Got error in Rev.AI receiver: sent 1000 (OK); then received 1000 (OK) Closing connection
main-telephony-app-1  | Terminating Rev.AI transcriber receiver

LLM

GPT-4

Transcription Services

RevAI

Synthesis Services

Microsoft Azure

Telephony Services

Twilio

Conversation Type and Platform

No response

Steps to Reproduce

Use this config to setup

telephony_server = TelephonyServer(
    base_url=BASE_URL,
    config_manager=config_manager,
    inbound_call_configs=[
        TwilioInboundCallConfig(
            url="/inbound_call",
            agent_config=ChatGPTAgentConfig(
                openai_api_key=OPENAI_API_KEY,
                model_name="gpt-4o",
                initial_message=BaseMessage(text="Hi, I am your customer support agent. How may i help you today?"),
                prompt_preamble="""You are a customer support agent. Your role is to provide resolution to a customer that is on call with you. You can make-up information to play along, to do this conversation. Just make it sound like a real conversation on telephone. Important: Generate short responses.""",
            ),
            twilio_config=TwilioConfig(
                account_sid=os.environ["TWILIO_ACCOUNT_SID"],
                auth_token=os.environ["TWILIO_AUTH_TOKEN"],
            ),
            synthesizer_config=AzureSynthesizerConfig(
                voice_name="en-GB-LibbyNeural",
                sampling_rate=DEFAULT_SAMPLING_RATE,
                audio_encoding=DEFAULT_AUDIO_ENCODING,
            ),
            transcriber_config=RevAITranscriberConfig.from_telephone_input_device(
            ),
        )
    ],
)

Expected Behavior

RevAI should transcribe and work, but it is not catching anything at all.

Screenshots

No response