VS Code CUDA Device IDs Extension

A small VS Code extension that shows CUDA device IDs in the status bar by reading the device_ids entry from your .devcontainer/docker-compose.yml. It's useful when you work with development containers or remote hosts that expose NVIDIA GPUs — the status bar shows which GPU device IDs are available so you can quickly reference them.

Quick overview (for users)

• What it shows: CUDA IDs: <ids> in the left status bar when device_ids are detected.
• Where it reads IDs from: .devcontainer/docker-compose.yml (supports nested locations such as deploy.resources.reservations.devices[*].device_ids).
• Default behavior: the extension is allowed to run on remote extension hosts by default. You can opt out via settings.
• Debug helper: Command Palette → CUDA Device IDs: Debug Check — shows remote name and whether device IDs were found.

Usage

1. Make sure your workspace contains .devcontainer/docker-compose.yml with a device_ids entry under a service (examples below).
2. The extension will automatically parse the file and show CUDA IDs: <ids> in the status bar.
3. If CUDA is not available on the host, the extension appends (No CUDA engine) to the status text.

Live GPU utilization menu

Click the status bar item (or run the command CUDA Device IDs: Show GPU Utilization) to open a small live menu that shows GPU utilization as short bars (e.g. ❚❚❚░░) and percent. The menu updates every 1 second and stops updating when you close it. Requires nvidia-smi to be available on the host.

Screenshot: example of the live GPU utilization menu:

Quick example of supported YAML (simple and nested examples):

Simple:

services:
   app:
      device_ids: ["0", "1"]

Nested (deploy -> resources -> reservations -> devices[*] -> device_ids):

services:
   app:
      deploy:
         resources:
            reservations:
               devices:
                  - driver: nvidia
                     capabilities: [gpu]
                     device_ids: ["2"]

Quick settings & requirements

• Requirements: Visual Studio Code and a workspace with .devcontainer/docker-compose.yml that contains device_ids (or nested devices entries).
• Enable/disable remotely: Setting cudaIds.showInRemoteHosts controls whether the extension runs on remote hosts (default: true).
  • Settings UI: File → Preferences → Settings and search for "CUDA Device IDs" or "Show In Remote Hosts".
  • Settings JSON: add "cudaIds.showInRemoteHosts": true to your settings.

Development (how to run locally)

1. Clone the repository and open it in VS Code:

git clone https://github.com/akkadhim/vscode-cuda-ids-extension.git
cd vscode-cuda-ids-extension
npm install

2. Press F5 in VS Code to run the extension in an Extension Development Host window.

Packaging & publishing

This repo includes a helper PowerShell script at scripts/publish.ps1 that automates version bumping (edits package.json), compiling, packaging and optionally publishing. Examples:

# package (no publish)
.\scripts\publish.ps1 -Bump none -PackageOnly

# bump patch, package and publish (prompts for PAT)
.\scripts\publish.ps1 -Bump patch -Publish

Notes:

• The helper script edits package.json but does not commit or push — commit/push manually from VS Code when you're ready.
• The script also writes package.json without a UTF-8 BOM to avoid packaging errors.
• If you prefer manual packaging:

npm run compile
npx vsce package

Contributing

Contributions are welcome. Feel free to open issues or pull requests.

License

MIT — see the LICENSE file for details.