Building audio/telephony/Virtual avatar agents with Livekit & LangGraph

Prerequisites

1. Follow the instructions in the reference documents to:
   • Install LiveKit and Twilio CLI
   • Configure the Twilio SIP trunk for inbound and outbound calls
   • Connect the Twilio SIP trunk to LiveKit
   • Set up LiveKit Cloud
2. Have an AI Agent ready (either a LangGraph agent hosted at a local/public URL or a LiveKit AI Agent)
3. Have all necessary API keys ready according to the .env.template file

Available agents

Agent	Description
quickstart_agent.py	A basic Livekit agent that can be used as a starting point
livekit_phone_calling_agent.py	A basic Livekit agent enhanced with telephony capabilities
langgraph_phone_calling_agent.py	A LangGraph agent enhanced with audio and telephony capabilities
langgraph_agent_avatar.py	A LangGraph agent enhanced with audio and AI Avatar

Instructions

Setup and Run

1. Create and activate virtual environment:

   uv venv && source .venv/bin/activate

2. Download model files (only need to do this once):

   uv run langgraph_phone_calling_agent.py download-files

3. Run the agent (could be any agents provided in this repository):

   uv run langgraph_phone_calling_agent.py dev

Testing Options

For Audio Agent and Avatar Agent

• Go to LiveKit Hosted Playground and connect to the room created in Step 1
• You can then talk with your agent

For Phone Calling Agent

• Use the following terminal command to start the phone call:

  lk dispatch create \
      --new-room \
      --agent-name my-telephony-agent \
      --metadata '{"phone_number": "+15105550123"}' # insert your own phone number here

Custom Langgraph plugins

Refer to temp_plugin/langchain

References

1. LiveKit Voice AI Guide
2. LiveKit Telephony Guide
3. Configuring SIP Trunk
4. Configuring Twilio Trunk