> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/moeru-ai/airi/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction to Airi

> An open-source, self-hosted AI companion platform that brings virtual characters to life with real-time interaction, gaming capabilities, and multi-platform support

## What is Airi?

Airi is a self-hosted AI companion and VTuber platform inspired by [Neuro-sama](https://www.youtube.com/@Neurosama). It allows you to create and interact with virtual characters powered by large language models, complete with voice chat, visual models (Live2D/VRM), and the ability to play games alongside you.

<Note>
  Unlike closed platforms, Airi gives you complete control over your digital companion. Everything runs on your own hardware, with no cloud dependencies unless you choose them.
</Note>

## Key Features

<CardGroup cols={2}>
  <Card title="Real-time Voice Chat" icon="microphone">
    Natural conversations with client-side speech recognition, voice activity detection, and support for multiple TTS providers including ElevenLabs.
  </Card>

  <Card title="Live2D & VRM Support" icon="user-circle">
    Bring your characters to life with full Live2D and VRM model support, including auto-blink, look-at tracking, and idle animations.
  </Card>

  <Card title="Game Integration" icon="gamepad">
    Your AI companion can play Minecraft and Factorio with you as a genuine teammate, not just a spectator.
  </Card>

  <Card title="Multi-Platform" icon="devices">
    Run on Web (WebGPU-powered), Desktop (Electron with native CUDA/Metal), or Mobile (iOS/Android via Capacitor).
  </Card>

  <Card title="40+ LLM Providers" icon="brain">
    Connect to OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Ollama, OpenRouter, and 40+ other providers through xsAI.
  </Card>

  <Card title="Persistent Memory" icon="database">
    In-browser database support with DuckDB WASM for context-aware conversations that remember past interactions.
  </Card>

  <Card title="Chat Platform Integration" icon="comments">
    Integrate with Discord voice channels, Telegram bots, and other platforms to bring your companion everywhere.
  </Card>

  <Card title="Extensible Architecture" icon="puzzle-piece">
    Plugin system powered by Model Context Protocol (MCP) for easy customization and integration.
  </Card>
</CardGroup>

## What Makes Airi Special?

Unlike other AI-driven VTuber projects, Airi was built from day one with web technologies:

* **WebGPU** for accelerated graphics
* **WebAudio** for audio processing
* **Web Workers** for parallel processing
* **WebAssembly** for performance-critical code
* **WebSocket** for real-time communication

<Warning>
  Worried about web technology performance?

  While the browser version showcases what's possible with modern web APIs, the desktop version (Stage Tamagotchi) natively supports NVIDIA CUDA and Apple Metal for maximum performance without complex dependency management.
</Warning>

### Platform Flexibility

Airi offers three deployment options to match your needs:

| Platform    | Name             | Best For                      | Key Features                                          |
| ----------- | ---------------- | ----------------------------- | ----------------------------------------------------- |
| **Web**     | Stage Web        | Quick access, no installation | WebGPU acceleration, PWA support, runs anywhere       |
| **Desktop** | Stage Tamagotchi | Power users, local models     | Native CUDA/Metal, screen capture, system integration |
| **Mobile**  | Stage Pocket     | On-the-go interaction         | iOS/Android apps, notifications, mobile-optimized UI  |

## Architecture Overview

Airi is built as a monorepo using pnpm workspaces with Vue 3, TypeScript, and a modular architecture:

```mermaid theme={null}
graph TD
    A[Stage Apps] --> B[Stage UI]
    B --> C[Core Modules]
    C --> D[xsAI Provider]
    C --> E[Audio Pipelines]
    C --> F[Memory System]
    C --> G[Character Models]
    
    H[Server Runtime] --> C
    I[Plugins] --> H
    J[Services] --> H
    
    D --> K[40+ LLM Providers]
    E --> L[Speech Recognition]
    E --> M[Voice Synthesis]
    F --> N[DuckDB/PGVector]
    G --> O[Live2D/VRM]
```

### Core Components

<AccordionGroup>
  <Accordion title="Stage UI" icon="window">
    Business components, composables, and stores shared across all three platforms. Houses provider definitions, orchestration modules, and reusable Vue components.

    Location: `packages/stage-ui/`
  </Accordion>

  <Accordion title="xsAI Integration" icon="brain-circuit">
    Unified interface for interacting with 40+ LLM providers. Similar to Vercel AI SDK but significantly smaller and more focused.

    Supported providers: OpenAI, Anthropic, Google Gemini, DeepSeek, Ollama, vLLM, SGLang, and many more.
  </Accordion>

  <Accordion title="Audio Pipelines" icon="waveform">
    Client-side speech recognition with voice activity detection, audio transcription through multiple providers (via unspeech), and text-to-speech synthesis.

    Location: `packages/pipelines-audio/`, `packages/audio/`
  </Accordion>

  <Accordion title="Memory System" icon="memory">
    Persistent conversation history and context using DuckDB WASM (in-browser) or PostgreSQL with pgvector (server-side). The Memory Alaya system (WIP) will provide advanced RAG capabilities.

    Location: `packages/drizzle-duckdb-wasm/`, `packages/memory-pgvector/`
  </Accordion>

  <Accordion title="Character Models" icon="person-rays">
    Full Live2D and VRM model support with automatic blink, look-at tracking, and idle animations powered by Three.js.

    Location: `packages/stage-ui-three/`, `packages/stage-ui-live2d/`
  </Accordion>

  <Accordion title="Server Runtime" icon="server">
    Powers bot integrations (Discord, Telegram, Minecraft, Factorio) and plugin execution. Supports WebSocket connections for remote control.

    Location: `packages/server-runtime/`, `packages/server-sdk/`
  </Accordion>
</AccordionGroup>

## Use Cases

<Steps>
  <Step title="Personal AI Companion">
    Create a virtual friend who remembers your conversations, understands your interests, and is always available to chat via voice or text.
  </Step>

  <Step title="Gaming Buddy">
    Play Minecraft or Factorio with an AI teammate that can build, gather resources, and strategize alongside you.
  </Step>

  <Step title="Virtual Streamer">
    Set up a VTuber that can interact with your Twitch/Discord community, respond to chat, and provide entertaining commentary.
  </Step>

  <Step title="Development Assistant">
    Integrate with your development workflow through VSCode extensions and MCP plugins for context-aware coding help.
  </Step>

  <Step title="Language Learning Partner">
    Practice conversations in different languages with immediate feedback and natural dialogue flow.
  </Step>
</Steps>

## Community and Ecosystem

Airi has spawned an entire ecosystem of sub-projects under [@proj-airi](https://github.com/proj-airi) and [@moeru-ai](https://github.com/moeru-ai):

* **[xsAI](https://github.com/moeru-ai/xsai)** - Universal LLM provider interface
* **[unspeech](https://github.com/moeru-ai/unspeech)** - Universal endpoint proxy for ASR and TTS
* **[MCP Launcher](https://github.com/moeru-ai/mcp-launcher)** - Easy MCP server management
* **[Velin](https://github.com/luoling8192/velin)** - Vue SFC + Markdown for LLM prompts
* **[AIRI Factorio](https://github.com/moeru-ai/airi-factorio)** - Factorio game integration
* **[Awesome AI VTuber](https://github.com/proj-airi/awesome-ai-vtuber)** - Curated list of AI VTuber projects

<Card title="Join the Community" icon="users" href="https://discord.gg/TgQ3Cu2F7A">
  Connect with other developers and users on Discord to share ideas, get help, and contribute to the project.
</Card>

## Technology Stack

<Tabs>
  <Tab title="Frontend">
    * **Framework**: Vue 3 with TypeScript
    * **Build Tool**: Vite 7
    * **State Management**: Pinia
    * **Routing**: Vue Router
    * **UI Components**: Reka UI (headless)
    * **Styling**: UnoCSS
    * **3D Graphics**: Three.js with TresJS
    * **Internationalization**: Vue I18n
  </Tab>

  <Tab title="Desktop">
    * **Framework**: Electron
    * **IPC**: @moeru/eventa (type-safe)
    * **Dependency Injection**: injeca
    * **Screen Capture**: electron-screen-capture
    * **Updates**: electron-updater
  </Tab>

  <Tab title="Mobile">
    * **Framework**: Capacitor
    * **Platforms**: iOS (Swift), Android (Kotlin)
    * **Native Features**: Local notifications, camera access
  </Tab>

  <Tab title="Backend">
    * **Runtime**: Node.js 23+
    * **Package Manager**: pnpm
    * **Monorepo**: Turborepo
    * **Database**: DuckDB WASM / PostgreSQL
    * **ORM**: Drizzle
    * **Schema Validation**: Valibot / Zod
  </Tab>
</Tabs>

## Project Status

Airi is under active development with regular updates. Check the [DevLogs](https://airi.moeru.ai/docs/en/blog/) for recent progress:

* Latest release: v0.8.5-beta.4
* Active contributors: Growing community
* License: MIT (open source)
* Production ready: Beta (use for personal projects)

<Warning>
  **Important**: Project AIRI has no official cryptocurrency or token. Beware of scams.
</Warning>

## Next Steps

Ready to get started with Airi?

<CardGroup cols={2}>
  <Card title="Quickstart" icon="rocket" href="/quickstart">
    Get your first AI companion running in 5 minutes
  </Card>

  <Card title="Installation Guide" icon="download" href="/installation">
    Detailed setup instructions for all three platforms
  </Card>

  <Card title="Configure Providers" icon="sliders" href="/features/llm-providers">
    Set up your preferred LLM provider and API keys
  </Card>

  <Card title="Import Characters" icon="user-plus" href="/features/character-models">
    Add Live2D or VRM models to customize your companion
  </Card>
</CardGroup>