Turn Any LangGraph Agent Into a Voice Agent in Minutes

Short summary

LangChain engineer Caroline di Vittorio shows how to add voice capabilities to existing LangGraph multi-agent systems using Pipecat, covering stateless graph design, STT/TTS/VAD handling, and complete LangSmith tracing integration. The tutorial includes a real customer-support demo with sub-agent routing, working code, and step-by-step walkthrough of the architecture changes needed.

• •Integrate LangGraph agents with Pipecat for voice without rebuilding
• •Key change: make graphs stateless to work with Pipecat's pipeline
• •LangSmith tracing captures audio, latency, costs across all layers

Generated with AI, which can make mistakes.