why Multilingual() model is not running in console ?

Thanks for reaching out, Adarsh. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

its not taking audio input from my microphone, and the error is not in my microphone because when i ran python agent.py start and connected in playground, it was working INFOlivekit.agentsstarting worker 2025-09-19 101726,672 - INFO livekit.agents - starting worker {"version": "1.2.8", "rtc-version": "1.0.13"} INFOlivekit.agentsstarting inference executor 2025-09-19 101726,672 - INFO livekit.agents - starting inference executor INFOlivekit.agentsinitializing process 2025-09-19 101726,702 - INFO livekit.agents - initializing process {"pid": 8740, "inference": true} DEBUGlivekit.agentsinitializing inference runner DEBUGlivekit.agentsinitializing inference runner 2025-09-19 101728,294 - DEBUG livekit.agents - initializing inference runner {"runner": "lk_end_of_utterance_multilingual", "pid": 8740, "inference": true} DEBUGlivekit.agentsinference runner initialized DEBUGlivekit.agentsinference runner initialized 2025-09-19 101733,586 - DEBUG livekit.agents - inference runner initialized {"runner": "lk_end_of_utterance_multilingual", "elapsed_time": 5.2913268999545835, "pid": 8740, "inference": true} DEBUGasyncioUsing proactor: IocpProactor INFOlivekit.agentsprocess initialized 2025-09-19 101733,587 - INFO livekit.agents - process initialized {"pid": 8740, "inference": true, "elapsed_time": 6.88} DEBUGasyncioUsing proactor: IocpProactor 2025-09-19 101733,587 - DEBUG asyncio - Using proactor: IocpProactor {"pid": 8740, "inference": true} INFOlivekit.agentsinitializing job runner 2025-09-19 101733,591 - INFO livekit.agents - initializing job runner {"tid": 20604} DEBUGasyncioUsing proactor: IocpProactor 2025-09-19 101733,592 - DEBUG asyncio - Using proactor: IocpProactor INFOlivekit.agentsjob runner initialized 2025-09-19 101733,592 - INFO livekit.agents - job runner initialized {"tid": 20604, "elapsed_time": 0.0} ERRORwhispey sdkFailed to extract configuration during setup: name 'telemetry_instance' is not defined 2025-09-19 101734,206 - ERROR whispey-sdk - Failed to extract configuration during setup: name 'telemetry_instance' is not defined INFOwhispey sdkSession data reference set for enhanced model detection 2025-09-19 101734,206 - INFO whispey-sdk - Session data reference set for enhanced model detection INFOwhispey sdkSession data HAS complete_configuration 2025-09-19 101734,207 - INFO whispey-sdk - Session data HAS complete_configuration INFOwhispey sdkStored telemetry instance reference with 0 spans 2025-09-19 101734,207 - INFO whispey-sdk - Stored telemetry instance reference with 0 spans DEBUGlivekit.agentshttp_session(): creating a new httpclient ctx 2025-09-19 101734,568 - DEBUG livekit.agents - http_session(): creating a new httpclient ctx [Audio] x (Realtek(R) Audio) [-70.00 dBFS] [------------------------------]DEBUGlivekit.agentsusing audio io:

ChatCLI

->

AgentSession

->

TranscriptSynchronizer

->

ChatCLI

2025-09-19 101734,570 - DEBUG livekit.agents - using audio io:

ChatCLI

->

AgentSession

->

TranscriptSynchronizer

->

ChatCLI

DEBUGlivekit.agentsusing transcript io:

AgentSession

->

TranscriptSynchronizer

->

ChatCLI

2025-09-19 101734,570 - DEBUG livekit.agents - using transcript io:

AgentSession

->

TranscriptSynchronizer

->

ChatCLI

[Audio] x (Realtek(R) Audio) [-98.99 dBFS] [------------------------------]INFOlivekit.agentsSTT metrics 2025-09-19 101739,637 - INFO livekit.agents - STT metrics {"audio_duration": 5.0} INFOlivekit.agentsSTT metrics 2025-09-19 101739,638 - INFO livekit.agents - STT metrics {"audio_duration": 5.0} [Audio] x (Realtek(R) Audio) [-97.96 dBFS] [------------------------------]INFOlivekit.agentsSTT metrics 2025-09-19 101744,639 - INFO livekit.agents - STT metrics {"audio_duration": 5.05} INFOlivekit.agentsSTT metrics 2025-09-19 101744,639 - INFO livekit.agents - STT metrics {"audio_duration": 5.05} [Audio] x (Realtek(R) Audio) [-96.13 dBFS] [------------------------------]INFOlivekit.agentsSTT metrics 2025-09-19 101749,734 - INFO livekit.agents - STT metrics {"audio_duration": 5.05} INFOlivekit.agentsSTT metrics

Hi there! I wanted to share some helpful information about the

Multilingual()

model and how it works with microphone audio. It doesn’t process raw audio directly but uses a Speech-to-Text (STT) plugin to convert the audio into text for analysis. Let’s go through some friendly steps to make sure everything runs smoothly: • Set Up an STT Plugin: You can use plugins like Deepgram or Gladia within the

AgentSession

to transcribe your microphone audio. Here’s a simple example:

from livekit.plugins import deepgram
from livekit.plugins.turn_detector.multilingual import MultilingualModel

session = AgentSession(
    turn_detection=MultilingualModel(),
    stt=deepgram.STT(model="nova-3", language="multi"),
)

• Check Language Settings: Make sure your STT plugin supports multilingual input by setting

language="multi"

. • Capture and Share Audio: Your app should capture microphone audio and publish it to the LiveKit room. For example:

audio_source = rtc.AudioSource(SAMPLE_RATE, NUM_CHANNELS)
local_audio_track = rtc.LocalAudioTrack.create_audio_track("microphone", audio_source)
await room.local_participant.publish_track(local_audio_track)

• Download Model Weights: Don’t forget to run

python agent.py download-files

so the model has the necessary inference weights ready. • Turn Off Conflicting Detection: If you’re using a Realtime LLM, it’s best to disable its built-in turn detection to avoid any conflicts. Following these steps should help get everything working perfectly. If you have any questions or need a hand with anything else, please feel free to ask—I’m here to help! Sources: Improving voice AI's turn detection with transformers | LiveKit turn detector plugin | LiveKit Docs | Building the all-in-one platform for voice AI agents | terminal-client/client.py | livekit-plugins/livekit-plugins-aws/livekit/plugins/aws/experimental/realti...