Skip to main content

Overview

Stage Tamagotchi is the native desktop application for AIRI, built with Electron. It provides the best performance and full access to native system capabilities including CUDA/Metal acceleration, screen capture, and system integrations.
Download the latest release from GitHub Releases

System Requirements

Minimum Requirements

  • Windows: Windows 10 64-bit or later
  • macOS: macOS 11.0 (Big Sur) or later
  • Linux: Ubuntu 20.04+ or equivalent
  • RAM: 4GB minimum, 8GB recommended
  • Storage: 2GB free space
  • GPU: Recommended for optimal performance
  • NVIDIA GPU: For CUDA acceleration (Windows/Linux)
  • Apple Silicon: For Metal acceleration (macOS)
  • Intel/AMD GPU: For general acceleration
  • RAM: 16GB for large models
  • SSD: For faster model loading

Installation

Download Pre-built Binaries

  1. Download AIRI-{version}-windows-x64-setup.exe from GitHub Releases
  2. Run the installer
  3. Follow the setup wizard
  4. Launch AIRI from Start Menu or Desktop shortcut
Windows may show a SmartScreen warning. Click “More info” → “Run anyway” if you trust the source.

Nix Package

For NixOS or Nix package manager users:
Or add to your flake.nix:

Building from Source

Prerequisites

  • Node.js 20+
  • pnpm 10+
  • Git
  • Platform-specific build tools:
    • Windows: Visual Studio 2019+ with C++ tools
    • macOS: Xcode 26+ (or Xcode 14+ for legacy icon format)
    • Linux: build-essential, libgtk-3-dev

Clone and Install

Development Mode

Start the development server:
This launches the Electron app with hot reload enabled.

Production Build

Architecture

Electron App Structure

IPC Communication

AIRI uses @moeru/eventa for type-safe IPC/RPC:

Native Features

GPU Acceleration

Automatic CUDA support on Windows/Linux with NVIDIA GPUs:
  • Accelerated model inference via candle
  • Real-time speech recognition
  • Faster model loading
  • Reduced CPU usage
CUDA support is automatically detected. No manual configuration needed.

Screen Capture

Native Integrations

  • System Tray: Minimize to tray
  • Notifications: Native system notifications
  • File System: Full file system access
  • Auto-Start: Launch on system startup
  • Deep Links: Custom URL protocol (airi://)
  • Keyboard Shortcuts: Global hotkeys

Auto-Updater

AIRI includes automatic update checking via electron-updater:

How It Works

  1. App checks for updates on startup
  2. Downloads updates in background
  3. Prompts user to restart when ready
  4. Installs update on restart

Configuration

Updates are published to GitHub Releases. Configuration in package.json:

Manual Update Check

Settings → System → General → Check for Updates

Development Workflow

Hot Reload

Both main and renderer processes support hot reload:

Debugging

Use Chrome DevTools (automatically opens):
  • Cmd/Ctrl + Shift + I: Toggle DevTools
  • Cmd/Ctrl + R: Reload
  • Vue DevTools available at __devtools__/

Testing

Limitations

Compared to the web version, the desktop app:
Advantages
  • Native CUDA/Metal GPU acceleration
  • Full file system access
  • System integrations (tray, notifications, etc.)
  • Better performance
  • Offline support without PWA limitations
  • Can run background services
Considerations
  • Larger download size (~200MB vs ~50MB web)
  • Requires installation
  • Platform-specific builds needed
  • Updates require restart

Troubleshooting

App Won’t Start

  1. Check system requirements
  2. Try deleting config: ~/.config/airi (Linux/macOS) or %APPDATA%/airi (Windows)
  3. Check logs: Help → Show Logs
  4. Run from terminal to see error messages

GPU Not Detected

If GPU is not detected:
  • Update GPU drivers
  • Check CUDA/Metal is installed
  • Verify GPU meets minimum requirements

IPC Errors

If you see IPC-related errors:
  1. Clear cache and restart
  2. Check main process logs
  3. Verify preload script is loading
  4. Check Eventa contract definitions

Technology Stack

  • Framework: Electron 33+
  • Renderer: Vue 3 + Vite
  • Build: electron-vite + electron-builder
  • IPC: @moeru/eventa
  • DI: injeca
  • State: Pinia
  • GPU: CUDA (NVIDIA) / Metal (Apple) / Vulkan (fallback)
  • Database: DuckDB (native)
  • Audio: Native audio APIs

Web Platform

Browser-based version with WebGPU

Mobile Platform

iOS and Android support