repository·main·Indexed 43 days ago
https://github.com/livekit-examples/python-agents-examplesA collection of runnable examples for building production-ready voice, video, and telephony agents using the LiveKit Agents framework. Includes a healthcare intake demo featuring an Anam lip-synced avatar, starter templates for Python and Node.js, and a React-based frontend using Shadcn/ui.
This is a Next.js frontend designed for AI voice assistants. It utilizes the LiveKit JavaScript SDK and React Components to implement core features including:
The application includes its own token server to manage session authentication.
The Web Voice Assistant is a Next.js starter template for LiveKit Agents. It provides a pre-built voice interface using the LiveKit JavaScript SDK and includes support for:
This template is designed to be used as a frontend for LiveKit Agents.
This project implements a Retrieval-Augmented Generation (RAG) system for a voice agent. It uses the following technologies and patterns:
Project Structure:
main.py: The core agent implementation.scrape_docs.py: Scraper for the target documentation site.build_rag_data.py: Orchestrates building the RAG database.rag_db_builder.py: Implementation of the database builder.rag_handler.py: Logic for processing RAG queries.data/: Storage for the vector database files.The xAI Patient Intake agent is a family-medicine front desk automation designed to handle complex medical office tasks in a single conversation. It can identify callers, book/move appointments, answer policy questions, collect clinical intake, and escalate emergencies.
Speech Pipeline (xAI end-to-end):
xai/stt-1xai/grok-4.3 (with reasoning_effort="none")xai/tts-1 (voice carina)Note: The clinic uses an in-memory fake database. It is not a medical device and does not provide medical advice.
This application is a Next.js-based frontend designed for AI voice assistants. It utilizes the LiveKit JavaScript SDK and React Components to provide a complete user interface. Key features include:
The LiveKit Drive-Thru Demo is a full-stack voice agent experience consisting of a Python-based voice agent and a Next.js dashboard. The agent acts as a drive-thru attendant, taking orders via voice, while the dashboard provides a real-time visual representation of the user's cart.
Repository Layout:
drive-thru-agent: A Python worker running the 'Kelly' persona. It manages ordering tools and sends checkout RPCs (Remote Procedure Calls) when the user completes their order.frontend: A Next.js 15 application that joins a LiveKit room, streams browser audio, and renders an animated order board and controls.The project is organized as follows:
app/: Next.js application routes and API.components/agents-ui/: Core Agents UI components.components/ai-elements/: AI-specific UI elements.components/app/: Business logic location. This folder manages application state, behavior, and the composition of Shadcn UI components.components/ui/: Primitive shadcn/ui components.Key Application Components in components/app:
session-view.tsx: Initializes the LiveKit session and renders the main UI (chat, media tiles, control bar).view-controller.tsx: Manages transitions between the welcome view and the active session view.welcome-view.tsx: The UI shown when not connected to a session.chat-transcript.tsx: Manages chat transcript transitions.tile-layout.tsx: Manages media tile layouts and transitions.The LemonSlice agent provides an interactive salary negotiation experience with the following capabilities:
Users can select from three distinct personalities, each with unique voices, avatars, and negotiation styles:
The project follows a standard Next.js App Router structure:
app/: Contains layouts, pages, and the API route for connection details.components/: Contains React components like CharacterPortrait, GameStatus, and SessionView.hooks/: Contains custom hooks for connection management, transcription, and debugging.lib/: Contains shared utilities and TypeScript types.public/portraits/: Stores portrait images for NPCs and agents.public/fonts/, postcss.config.js, tailwind.config.ts: Styling and theming configuration.The project is organized into the following directories:
| Directory | Description |
|---|---|
battleground-agent | Agent 1 (Baseline): Uses AssemblyAI + GPT-4.1-mini + Inworld TTS. Contains the dispatch RPC used to spawn other agents. |
battleground-agent-2 | Agent 2 (Alt Stack A): Uses Deepgram + Gemini + Cartesia. |
battleground-agent-3 | Agent 3 (Alt Stack B): Uses Deepgram + Moonshot + Rime. |
frontend | Next.js dashboard for real-time metrics, transcripts, and dispatch controls. |
The project is organized to separate UI primitives, AI-specific elements, and application business logic:
app/: Next.js application routes and API endpoints.components/agents-ui/: Core Agents UI components.components/ai-elements/: Specialized AI components.components/app/: Business logic layer. This folder manages application state, behavior, and the composition of Shadcn UI components. Key files include:session-view.tsx: Initializes the LiveKit session and renders the main UI (chat, media tiles, control bar).view-controller.tsx: Manages transitions between the welcome screen and the active session.welcome-view.tsx: The UI shown before a session is connected.chat-transcript.tsx: Handles chat transcript transitions.tile-layout.tsx: Manages media tile layouts and transitions.