Hey there! Just a heads-up—LiveKit doesn’t come with a built-in feature to add an LLM router between the LLM and TTS for deciding on the fly whether to use TTS. But no worries, you can totally make this happen by customizing the agent pipeline. Here’s a simple way to do it: 1. Create your own custom

Agent

class. 2. Override the

tts_node

method to add some logic that checks if TTS should run based on what the LLM outputs. 3. If you want, you can also tweak the

llm_node

method to help prepare or signal this decision. 4. Then, return either a modified or empty stream from

tts_node

to control the audio synthesis. This way, you can flexibly skip TTS when certain conditions are met. If you want, I’m here to help you walk through the implementation—just let me know! Sources: livekit-agents/livekit/agents/voice/generation.py | Pipeline nodes and hooks | LiveKit Docs