Vitrine

A touch-optimized system monitoring dashboard for the HYTE Y70 Touch Infinite display (682x2560), built with Tauri v2, React 18, D3.js, and Tailwind CSS.

Vitrine replaces the Windows-only HYTE Nexus software with a native Linux application that provides real-time hardware telemetry, animated visualizations, and an optional AI advisor — all designed for the Y70's vertical touchscreen.

Vitrine is French for "glass display case" — a reference to the Y70's glass panel that showcases your hardware.

Features

• Real-time telemetry — CPU, GPU, RAM, NVMe, network, fans via sysfs
• D3.js visualizations — Arc gauges, radial progress rings, sparklines, animated bars, history graphs
• Configurable pages — Multi-page carousel with swipe navigation
• Widget system — 7 built-in widgets, configurable layout via TOML
• AI System Advisor — Optional LM Studio integration for ambient hardware insights
• Dark/light theming — Accent color picker, glow effects, animation controls
• Settings drawer — Swipe down from top edge to configure everything
• Config hot-reload — Edit config.toml and changes apply live

Quick Start (Bazzite / Fedora)

1. Install dependencies

# System libraries
sudo dnf install webkit2gtk4.1-devel gtk3-devel libappindicator-gtk3-devel librsvg2-devel pango-devel

# Rust (if not already installed)
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source ~/.cargo/env

# Node.js + pnpm (if not already installed)
sudo dnf install nodejs
npm install -g pnpm

# Tauri CLI
cargo install tauri-cli --version "^2"

2. Clone and build

git clone https://github.com/AppSprout-dev/vitrine.git
cd vitrine
pnpm install
pnpm tauri build

The binary will be at src-tauri/target/release/vitrine.

3. Install

chmod +x install.sh
./install.sh

This will:

1. Copy the binary to ~/.local/bin/vitrine
2. Create a default config at ~/.config/vitrine/config.toml
3. Install and enable a systemd user service (auto-starts on login)
4. Auto-detect the Y70 display and configure KDE window rules (if applicable)

4. Verify it's running

systemctl --user status vitrine

To restart after config changes:

systemctl --user restart vitrine

Configuration

Config file location: ~/.config/vitrine/config.toml

[general]
polling_interval_ms = 1000    # Sensor polling rate (250-5000)
mock_sensors = false          # Force mock data (for testing)
auto_start = true
target_display = "auto"       # Auto-detect Y70, or specify output name

[theme]
mode = "dark"                 # "light" or "dark"
accent_color = "#60A5FA"      # Any hex color
background_opacity = 0.85     # 0.5-1.0
enable_glow_effects = true
enable_animations = true

[temperature_thresholds]
cpu_warn = 75
cpu_critical = 90
gpu_warn = 80
gpu_critical = 95

[[pages]]
name = "System"

[[pages.widgets]]
widget_type = "cpu_monitor"
position = { col = 0, row = 0 }
size = { cols = 2, rows = 2 }

[[pages.widgets]]
widget_type = "gpu_monitor"
position = { col = 0, row = 2 }
size = { cols = 2, rows = 2 }

[[pages.widgets]]
widget_type = "ram_usage"
position = { col = 0, row = 4 }
size = { cols = 1, rows = 1 }

[[pages.widgets]]
widget_type = "nvme_storage"
position = { col = 1, row = 4 }
size = { cols = 1, rows = 1 }

Available Widgets

Type	Description	Default Size
cpu_monitor	CPU temp, utilization, frequency, per-core bars	2x2
gpu_monitor	GPU temp, utilization, VRAM, power draw	2x2
ram_usage	System + swap memory radial gauges	1x1
nvme_storage	Drive temp, read/write throughput	1x1
network_monitor	Interface RX/TX speeds	1x1
media_player	MPRIS2 now-playing info	2x2
system_advisor	AI-powered hardware insights (requires LM Studio)	2x1

AI System Advisor (Optional)

The System Advisor widget connects to a locally-running LM Studio instance to provide ambient hardware insights. It periodically analyzes your telemetry and surfaces observations like thermal trends, efficiency tips, or anomaly alerts.

Vitrine works perfectly fine without it. The advisor is disabled by default if LM Studio isn't running.

Setup

1. Install LM Studio and load a small model (we recommend ibm/granite-4-h-tiny for fast, clean responses — avoid "thinking" models like Qwen 3.5 which waste tokens on chain-of-thought)
2. Start the LM Studio local server (default: http://localhost:1234)
3. Add to your config.toml:

[advisor]
enabled = true
api_endpoint = "http://127.0.0.1:1234/v1/chat/completions"
model = "ibm/granite-4-h-tiny"   # Or leave empty to auto-detect
interval_seconds = 90             # How often to generate insights

The advisor will start generating insights after collecting 2 minutes of telemetry data. Tap an insight in the expanded view to get a more detailed explanation.

Advisor Config Reference

Field	Default	Description
enabled	true	Enable/disable the advisor
api_endpoint	http://127.0.0.1:1234/v1/chat/completions	LM Studio API URL
model	"" (auto-detect)	Model ID to use, or empty to pick first available
interval_seconds	90	Seconds between insight generation
min_data_seconds	120	Minimum telemetry data before first insight
request_timeout_seconds	180	HTTP timeout (increase for slow models)
max_insight_history	50	Max insights to keep in memory
system_prompt	(built-in)	Custom system prompt for the model

Development

Prerequisites

• Rust 1.70+, Node.js 18+, pnpm
• Tauri CLI v2: cargo install tauri-cli --version "^2"

Dev Server

pnpm install
pnpm tauri dev

This starts Vite (frontend hot-reload) + Rust backend. On non-Linux systems, mock sensor data is used automatically.

Force mock sensors on Linux:

VITRINE_MOCK_SENSORS=1 pnpm tauri dev

Checks

pnpm lint                      # ESLint
npx tsc --noEmit               # TypeScript type check
cd src-tauri && cargo clippy    # Rust lints

Architecture

src-tauri/src/
├── lib.rs              # App setup, polling loop, config watcher
├── commands.rs         # IPC handlers (sensors, config, advisor)
├── config.rs           # TOML config schema with serde
├── sensors/
│   ├── mod.rs          # SensorProvider trait, SensorData structs
│   ├── cpu.rs          # Linux sysfs reader
│   └── mock.rs         # Mock data generator (auto on non-Linux)
└── advisor/
    ├── engine.rs       # Advisor thread, scheduling, event detection
    ├── buffer.rs       # Telemetry ring buffer with rolling stats
    ├── client.rs       # HTTP client for LM Studio API
    ├── prompt.rs       # Prompt construction, category classification
    └── events.rs       # Thermal/load/anomaly event detection

src/
├── App.tsx                         # Root: pages, theme, settings
├── hooks/
│   ├── useSensorData.ts            # Tauri event subscription
│   ├── useConfig.ts                # Config read/write via IPC
│   ├── useAdvisorInsights.ts       # Advisor event + IPC
│   ├── useTheme.ts                 # DOM theme application
│   └── useHistoryBuffer.ts         # Rolling data buffer
├── components/
│   ├── layout/                     # PageCarousel, WidgetGrid, PageIndicator
│   ├── widgets/                    # CpuMonitor, GpuMonitor, SystemAdvisor, etc.
│   ├── visualizations/             # ArcGauge, RadialProgress, Sparkline, etc.
│   └── settings/                   # SettingsDrawer, GeneralTab, AppearanceTab
└── lib/
    ├── types.ts                    # TypeScript interfaces (snake_case)
    ├── constants.ts                # Panel dimensions, colors
    ├── widgetRegistry.ts           # Widget type -> component mapping
    └── formatters.ts               # Display value formatters

Contributing

See CONTRIBUTING.md for setup instructions and guidelines.

License

GPL v3