I'm happy to help you with LiveKit's support for using

MultimodalAgent

alongside the Gemini Live API! You can use the syntax

const agent = new multimodal.MultimodalAgent({ model, fncCtx, chatCtx: initialContext });

in the Node.js SDK, and it works well. Just make sure your

model

is an instance of

google.beta.realtime.RealtimeModel

, set up with Gemini-specific settings like

instructions

,

voice

, and

api_key

. While the

fncCtx

and

chatCtx

parameters are optional, they’re really helpful if you want your agents to be stateful and functional. If you’d like to dive deeper, I recommend checking out the Gemini integration guide and the Node.js multimodal agent example. They’re great resources to guide you through the process!