CPU/GPU Tray Monitor

A lightweight system-resource monitor that lives in the GNOME top bar. The panel icon shows your current CPU temperature as a color-coded number, and clicking it opens a menu with the full breakdown: CPU and GPU temperature, utilization, and RAM / VRAM usage. It refreshes on a configurable interval (2 seconds by default).

Written in Rust with no GTK dependency — the tray is a pure StatusNotifierItem, CPU/RAM come from sysinfo, GPU metrics come from a small per-vendor layer (NVIDIA via NVML, AMD via sysfs, Intel best-effort), and the icon text is drawn directly to a bitmap with ab_glyph.

Screenshot

Features

• Color-coded CPU temperature in the panel: green when cool, amber when warm (>= 65 C), red when hot (>= 80 C).
• Click-to-open menu showing:
  • CPU temperature and utilization
  • Available RAM, total RAM, and RAM usage %
  • GPU temperature and utilization
  • Available VRAM, total VRAM, and VRAM usage %
• Multi-vendor GPU support, auto-detected at startup:
  • NVIDIA — full temp / utilization / VRAM via NVML
  • AMD — full temp / utilization / VRAM via sysfs (no extra tooling)
  • Intel — temperature only where exposed (see Limitations)
• Configurable via a TOML file — refresh interval and icon/font sizing.
• Degrades gracefully: a missing GPU, missing config file, or unreadable sensor falls back to sensible defaults rather than crashing.

Requirements

• Linux with a StatusNotifierItem host. On GNOME that means the AppIndicator and KStatusNotifierItem Support extension must be installed and enabled — without it the tray has nowhere to appear and the app exits on startup with an error.
• A GPU for the GPU metrics (CPU, RAM, and the panel icon work regardless):
  • NVIDIA: NVML is loaded at runtime from the driver you already have (libnvidia-ml); no extra package needed.
  • AMD: read straight from the kernel's amdgpu sysfs interface; nothing to install.
  • Intel: temperature only, and only on GPUs that expose it (discrete Arc); integrated GPUs usually don't. See Limitations.
• Rust toolchain (stable) to build — install via rustup.
• A font file at assets/DejaVuSans.ttf, embedded into the binary at compile time. On Ubuntu, copy the system copy into place (see Building).

Building

Clone the repo, then build a release binary:

cargo build --release

The optimized executable is written to target/release/resource-monitor

Running

Launch it directly:

cargo run --release
# or run the built binary:
./target/release/resource-monitor

The icon appears in the top bar. Click it for the full menu; Quit stops the app (or press Ctrl+C in the terminal). Because it polls forever, it won't exit on its own — that's expected.

Start automatically on login

Create ~/.config/autostart/resource-monitor.desktop:

[Desktop Entry]
Type=Application
Name=CPU/GPU Tray Monitor
Exec=/home/YOURUSER/.local/bin/resource-monitor
X-GNOME-Autostart-enabled=true

Copy the binary to ~/.local/bin/ first, and use the full absolute path in Exec= (autostart doesn't reliably use your shell PATH).

Configuration

Settings live in a TOML file at:

~/.config/resource-monitor/config.toml

The file is optional — if it's missing, or a field is omitted, or it fails to parse, the app falls back to the built-in defaults, so it never blocks startup. A parse error prints a warning and continues with defaults.

Available keys (shown with their defaults):

interval_secs = 2     # how often to refresh, in seconds
icon_width    = 32    # tray icon width in pixels
icon_height   = 32    # tray icon height in pixels
font_px       = 32.0  # temperature text size in pixels

A copy of this lives in config.example.toml in the repo — copy it to the path above and edit to taste.

A few things remain hardcoded in the source (change them there if needed):

• Temperature color thresholds — temp_color() in src/tray.rs.
• CPU temperature sensor selection — see How it works.

How it works

• config.rs defines the Config struct (with serde defaults) and loads it once via a OnceLock, so every module can read the same settings through config() without passing an object around. Reads ~/.config/resource-monitor/config.toml, falling back to defaults on any problem.
• metrics.rs holds the long-lived handles (sysinfo::System, sysinfo::Components, and a Gpu) and exposes read() returning a Readings snapshot. CPU temperature is matched robustly: it looks for a component label containing "Package id" (Intel) or "Tctl" (AMD), and if neither is present or returns a value, falls back to the hottest Core* reading.
• gpu.rs abstracts the GPU behind a Gpu enum (Nvidia / Amd / Intel / None), chosen once at startup: it tries NVML first, then scans /sys/class/drm/card* by PCI vendor id. Each backend returns the same GpuReadings (temp, utilization, memory used/total). The NVIDIA Device is fetched fresh each read because it borrows from the Nvml and can't be stored alongside it.
• render.rs draws the temperature string into an ARGB32 bitmap sized from the config. The buffer is built as straight RGBA, then each pixel is rotated one byte ([R,G,B,A] -> [A,R,G,B]) because ksni expects ARGB32 in network byte order.
• tray.rs implements ksni::Tray: icon_pixmap() renders the current CPU temp, and menu() builds the full read-out.
• main.rs spawns the tray, then loops: sleep for the configured interval, read fresh metrics, and handle.update(...), which makes ksni re-read both the icon and the menu.

Known limitations

• GPU coverage varies by vendor. NVIDIA and AMD report full temp / utilization / VRAM. Intel reports temperature only where the hardware exposes it (discrete Arc); integrated Intel GPUs generally share the CPU package and expose no separate temperature, and Intel utilization/VRAM aren't read at all — those require the i915/xe performance counters (PMU), which need elevated privileges and are out of scope here.
• The panel icon shows a single number. GNOME renders tray icons as a square via St.Icon, so a wider multi-value readout isn't possible through this path; the CPU temp is the headline and everything else lives in the menu.
• No hover tooltips. The GNOME AppIndicator extension doesn't implement StatusNotifierItem tooltips, which is why all detail is in the click menu.
• Memory is reported in decimal GB (/ 1,000,000,000), so values are slightly higher than the GiB figures shown by tools like free -h.

Project structure

src/
├── main.rs      # entry point: load config, spawn tray, poll loop
├── config.rs    # TOML config (serde) loaded once via OnceLock
├── metrics.rs   # data collection (CPU/RAM via sysinfo, GPU via gpu.rs)
├── gpu.rs       # vendor detection + per-vendor GPU backends
├── render.rs    # draws the temperature text into the tray icon
└── tray.rs      # ksni Tray impl: icon + menu
assets/
└── DejaVuSans.ttf     # embedded font (add manually)
config.example.toml
Cargo.toml

License

MIT License