are there any toggle or parameter available that can help improve latency for the voice bot?

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

what is fast-preresponse?

Fast-preresponse is like having a super-quick, attentive friend who immediately lets you know they're listening while they're thinking through their full response. It's a clever technique in LiveKit voice agents that makes conversations feel smooth and natural. • Imagine a lightweight, speedy AI that quickly says something like "Great point!" or "Hmm, let me think about that" the moment you finish speaking. • At the same time, another AI is working hard to craft a detailed, thoughtful answer just for you. • These quick acknowledgments are just a friendly nod – they don't stick around in the conversation, they're just there to keep things feeling warm and connected. • The goal? To make sure you never feel like you're talking into a void, and that someone (or something) is always right there with you, ready to help. By using this dual-AI approach, fast-preresponse creates a conversation that feels more like chatting with a friend than waiting for a computer to respond. Sources: examples/voice_agents/fast-preresponse.py | Live Conversations with AI using ChatGPT and WebRTC

Here on using this code agent is speaking twice how we can overcome this? class IVRNavigatorAgent(Agent): """ Agent 1: Listens for the IVR and uses the FAST LLM for a quick response. """ def __init__(self, *, fast_llm: llm.LLM): # super().__init__(instructions="You are an IVR navigator. Wait for the user to speak.") super().__init__(instructions=" ") self._fast_llm = fast_llm self._request_made = False async def on_user_turn_completed(self, turn_ctx: ChatContext, new_message: ChatMessage): if self._request_made: return self._request_made = True logger.info("IVR has spoken. Generating a fast initial response.") prompt = ( "You are an AI assistant calling a pharmacy. When the IVR answers, your only job is to say the following sentence exactly: " "'Hello, I would like to speak with a human agent to collect some information regarding a patient. Could you please connect me?'" ) initial_message = llm.ChatMessage(role="system", content=[prompt]) chat_ctx = llm.ChatContext(items=[initial_message]) stream = self._fast_llm.chat(chat_ctx=chat_ctx) await self.session.say(stream.to_str_iterable()) await self.request_completed() @function_tool() async def request_completed(self) -> Agent: logger.info("Initial request made. Handoff -> WaitingAgent") return WaitingAgent(chat_ctx=self.session._chat_ctx, fast_llm=self._fast_llm)

I don't have the answer you're looking for. You could also try asking your question: • in one of the other Slack channels or • to https://deepwiki.com/livekit/livekit_composite which is trained on all LiveKit source code If you find the answer, please post it here to help others!