Name: 'SavicLabs GPU Monitor'
Author: 'SavicLabs'

Real-time GPU monitor in your VS Code status bar. Shows VRAM, temperature, utilization — click for detailed panel with process list and history charts.

Features

Status bar widget — VRAM usage, temperature, utilization at a glance
Detail panel — per-GPU breakdown, VRAM bar, process list, 5-minute history graph
Multi-GPU — cycle through GPUs, or see all in the panel
Smart alerts — warns when VRAM exceeds 85% or GPU runs hot
Zero config — auto-detects NVIDIA, AMD, Apple Silicon, Intel Arc
Extension API — other extensions can query GPU state

Supported Hardware

Vendor	Linux	Windows	macOS
NVIDIA	✅ nvidia-smi + /proc fallback	✅ nvidia-smi	—
AMD	✅ rocm-smi	✅ WMI	—
Apple Silicon	—	—	✅ system_profiler
Intel Arc / UHD / Iris	✅ intel_gpu_top	✅ WMI	—

Usage

Once installed, the GPU status appears in the status bar:

GPU RTX 3060 · 8.2GB/12GB · 52°C

Click the status bar item to open the detail panel.

Right-click → toggle on/off.

Commands

Command	Description
SavicLabs GPU: Open Detail Panel	Open the per-GPU breakdown panel
SavicLabs GPU: Toggle Monitor	Enable/disable the status bar monitor
SavicLabs GPU: Refresh	Force immediate GPU stats refresh

Settings

Setting	Default	Description
`savicLabsGpu.enabled`	`true`	Enable the monitor
`savicLabsGpu.pollIntervalMs`	`2000`	Poll interval in ms
`savicLabsGpu.alertVramPercent`	`85`	VRAM % that triggers warning
`savicLabsGpu.alertTempCelsius`	`85`	Temp °C that triggers warning
`savicLabsGpu.statusBarFormat`	`GPU ${name} · ${vramUsed}/${vramTotal} · ${temp}°C`	Status bar format string
`savicLabsGpu.showUtilization`	`false`	Show GPU util % in status bar

Status Bar Variables

Variable	Example
`${name}`	RTX 3060
`${vramUsed}`	8.2GB
`${vramTotal}`	12GB
`${vramPercent}`	68%
`${temp}`	52
`${util}`	45

Extension API

Other extensions can query GPU state:

const gpu = vscode.extensions.getExtension('SavicLabs.savic-labs-gpu');
const api = await gpu?.activate();
api.getGpuStats();           // GpuStats[]
api.getAvailableVram();      // MiB free across all GPUs
api.canFitModel(20000);      // Can a 20GB model fit?