livekit-examples/python-agents-examples

What it solves

This repository provides a comprehensive library of runnable examples for developers building voice, video, and telephony agents. It bridges the gap between basic API documentation and production-ready implementations by demonstrating real-world patterns for AI agents that can hear, speak, see, and interact with external systems.

How it works

The project is organized into two main sections: focused, single-concept demos (in docs/examples/) and full-stack, production-style applications (in complex-agents/). It leverages the LiveKit Agents framework to orchestrate the pipeline of Speech-to-Text (STT), Large Language Models (LLMs), and Text-to-Speech (TTS), while integrating with various providers like OpenAI, Anthropic, Google Gemini, and Deepgram.

Who it’s for

Developers looking to implement AI voice assistants, automated phone systems (SIP), multimodal agents with vision capabilities, or animated digital avatars into their applications.

Highlights

• Diverse Agent Types: Includes examples for telephony (SIP), real-time vision, and animated avatars (Hedra, Tavus).
• Multi-Agent Orchestration: Demonstrates how to switch between specialized agents mid-call (e.g., medical triage or personal shopping).
• Pipeline Customization: Examples for intercepting and modifying the STT $\rightarrow$ LLM $\rightarrow$ TTS flow for content filtering or translation.
• External Integrations: Support for RAG (vector search), Home Assistant, and Model Context Protocol (MCP) clients.
• Full-Stack Demos: Complete applications including React frontends, such as a voice-driven D&D RPG and a deep research agent.