Thanks — I’ll help with that. Short answer: LiveKit does not have a direct Qwen realtime plugin. Qwen realtime models are not available in LiveKit Inference. The recommended approach is to use an external provider that hosts Qwen (Cerebras is mentioned as supporting open models like Qwen) and connect it to LiveKit via the OpenAI-compatible plugin. What you can do next: 1. Confirm where your Qwen model is hosted (e.g., Cerebras). 2. If using Cerebras, configure LiveKit Agents to use the OpenAI plugin with Cerebras credentials and the model name. Example pattern:

from livekit.plugins import openai

session = AgentSession(
    llm=openai.LLM.with_cerebras(
        model="your-qwen-model-on-cerebras",
    ),
)

1. Test for the behaviors you need (text-only LLM calls vs. realtime speech modes). Note: realtime speech-to-speech features available for OpenAI/xAI may not be available for Qwen via this path. If you’d like, tell me: • where your Qwen model is hosted, and • whether you need text-only LLM responses or realtime speech—so I can give step-by-step setup guidance. Relevant docs: Realtime models overview, Cerebras integration, and Qwen LLM pages. Sources: Realtime models overview | LiveKit Documentation | Cerebras and LiveKit | LiveKit Documentation | Qwen LLM | LiveKit Documentation | HIPAA Eligible Products and Services | LiveKit | OpenAI Realtime API plugin guide | LiveKit Documentation, with interactive elements