Agent Definition with LiveKit

Deploy real-time voice and video agents using LiveKit's open-source WebRTC infrastructure.

Features

• Real-Time Communication: Ultra-low latency voice and video
• Scalable Infrastructure: Production-ready WebRTC SFU
• Agent Framework: Built-in support for AI agents with voice
• Multi-Modal: Combine voice, video, and data streams

Prerequisites

Infrastructure

Requires Docker for running LiveKit server locally, or use LiveKit Cloud.

# Add lattice user to Docker group
sudo adduser lattice docker
sudo systemctl restart lattice

API Keys

Set your LiveKit credentials:

export LIVEKIT_URL="wss://your-livekit.livekit.cloud"
export LIVEKIT_API_KEY="your-api-key"
export LIVEKIT_API_SECRET="your-api-secret"

Architecture

This agent definition provisions:

• LiveKit server (local or cloud)
• Python environment with livekit-agents SDK
• Pre-configured voice pipeline (STT, LLM, TTS)
• WebRTC infrastructure for real-time audio

Note: Edit the Terraform to customize the LiveKit agent pipeline and voice models.

Getting Started

1. Create a new workspace from this template
2. Configure your LiveKit credentials
3. Deploy voice-enabled AI agents

Built with LiveKit - YC W21