livekit-voice-agent

What problem does it solve?

This Skill provides a ready-to-use pattern for building production-ready LiveKit voice agents that can hand off conversations between specialized agents and escalate to humans when needed. It also preserves conversation context across handoffs and supports scalable deployment patterns.

Core Features & Use Cases

• Multi-Agent Workflows: Chains IntroAgent, SpecialistAgent, and EscalationAgent with shared state and configurable handoffs.
• Intelligent Handoffs: Function tools return new agent instances to transfer control, with clear transition messages.
• Context Preservation: Shared user data persists across handoffs and across agents via chat_ctx.
• Production-Ready Patterns: Templates, prewarming, testing scaffolds, Docker deployment, and observability hooks.
• Extensibility: Easy to add new agents, tools, and custom workflows; supports hub-spoke, escalation, and bidirectional handoffs.
• Use Cases: Real-time customer support, sales routing, and complex information gathering with escalations.

Quick Start

• Run the quick start script to scaffold a working project: scripts/quickstart.sh livekit-voice-agent
• Install dependencies with uv sync
• Copy .env.example to .env.local and configure LiveKit, OpenAI, and STT/TTS providers
• Start the agent with: uv run python src/agent.py start