AIRI is an open‑source project that recreates the Neuro‑Sama AI VTuber using a stack of web technologies—WebGPU, WebAudio, Web Workers, WebAssembly, and WebSocket—for rendering, animation, and communication, while the desktop build falls back to native GPU backends (CUDA on NVIDIA, Metal on Apple) via the HuggingFace‑candle integration to avoid reliance on browsers alone. It connects to a variety of LLM APIs through the xsai abstraction layer, supporting providers such as OpenAI, Anthropic Claude, Google Gemini, xAI Groq, Ollama, vLLM, SGLang, and many others. The system loads VRM and Live2D models, drives their animations (auto‑blink, gaze‑following, idle eye movement), and accepts audio input from the browser or Discord, performing client‑side speech recognition and voice activity detection before generating responses with multi‑provider TTS (ElevenLabs, Azure Speech, OpenAI‑compatible, Alibaba Cloud, Kokoro). Memory is handled by in‑browser databases (DuckDB WASM, pglite) and an experimental Memory Alaya module. AIRI can play games such as Minecraft and Factorio (WIP), with planned support for Kerbal Space Program and Helldivers 2, and offers chat bridges to Telegram and Discord. Installation is provided via winget, Scoop, Homebrew Cask, pnpm, Nix, and Capacitor for mobile/PWA builds.