Thanks — I understand you tried to deploy a new agent from the UI but the realtime pipeline only lets you pick the Xgrok model and you want to use an OpenAI realtime model instead. I can help you switch it — here are the steps and some checks to make sure OpenAI appears as an option. What to try in the UI 1. Open the Agent Builder for the agent you’re creating. 2. When you reach the model selection step, change the model type to Realtime API (this selects realtime LLMs rather than the default Grok/Xgrok option). 3. After selecting Realtime API, choose OpenAI Realtime API from the model list. If OpenAI is not visible in the model list — quick checks • Make sure your project has an OpenAI API key configured in the environment used by the dashboard. The UI will only surface OpenAI if an

OPENAI_API_KEY

is available for the project. • Confirm the key is valid and has the necessary permissions for realtime usage. How to add the OpenAI key (if needed) • Add

OPENAI_API_KEY=your_openai_api_key_here

to the project environment variables used by the LiveKit dashboard (or the agent’s environment). Once set, refresh the Agent Builder UI and the OpenAI Realtime option should appear. Advanced / alternative: set model in configuration or CLI • If the UI still doesn’t let you pick it, you can explicitly set OpenAI in the agent configuration (or via CLI). Example for an AgentSession:

from livekit.plugins import openai

session = AgentSession(
    llm=openai.realtime.RealtimeModel(voice="alloy"),
)

• Or deploy with the CLI and ensure your agent config references the OpenAI RealtimeModel. If you’d like, tell me: • Which project you’re using (or whether you already added an OPENAI_API_KEY), and • Whether you’re working in the Cloud dashboard or self-hosted UI. With that info I’ll guide you step-by-step or point out the exact place to add the API key. Sources: OpenAI Realtime API plugin guide | LiveKit Documentation