llama-swap-status
Display llama-swap status in the VS Code status bar, including GPU temperature, token generation speed, and processed token counts.
Features
- GPU Temperature: Real-time GPU temperature display in the status bar
- Model Status: Optionally shows the active model name and status (ready, starting, stopping, etc.)
- Token Speed: Displays aggregate token generation speed across active slots (t/s)
- Processed Tokens: Shows prompt tokens processed during prefill, then prompt processed plus generated tokens during generation
- Detail Panel: Click any status bar item to open a detailed monitoring panel
- Automatic Reconnection: Detects llama-swap when it becomes available after startup
Status Bar Behavior
The extension shows four optional items on the right side of the VS Code status bar:
| Item |
Display |
Data source |
| GPU |
Temperature for each GPU |
llama-swap /api/performance |
| Model |
Active model name and status |
llama-swap /v1/models |
| Speed |
Aggregate generation speed across active slots |
llama-server /slots |
| Processed Tokens |
Prompt tokens processed during prefill, then prompt processed plus generated tokens |
llama-server /slots |
Model information is hidden by default to preserve status bar space. Processed token information follows the most active slot, while generation speed is aggregated across all active slots. When llama-swap is unavailable, the status bar items remain hidden and reappear automatically after the server becomes available.
Requirements
- A running llama-swap instance (default:
http://localhost:8080)
Extension Settings
This extension contributes the following settings:
llamaSwap.url: URL of the llama-swap server (default: http://localhost:8080)llamaSwap.refreshInterval: Polling interval in milliseconds (default: 3000)llamaSwap.slotsPollInterval: Slot polling interval in milliseconds (default: 1000)llamaSwap.showGpuInfo: Show GPU temperature in status bar (default: true)llamaSwap.showModelInfo: Show active model name and status in status bar (default: false)llamaSwap.showSpeedInfo: Show token speed in status bar (default: true)llamaSwap.showPromptProgress: Show processed token counts in status bar (default: true)
The llamaSwap.statusBarPosition setting is currently reserved for future use; status bar items are displayed on the right side.
Commands
llama-swap-status: Open Monitor Panel - Open detailed monitoring webviewllama-swap-status: Open Web UI - Open llama-swap web interfacellama-swap-status: Toggle Status Bar - Toggle status bar visibilityllama-swap-status: Refresh - Manually refresh connection and data
Development
npm install
npm run compile
code . # Then press F5 to debug
License
MIT
Release Notes
0.1.0
Initial release with GPU monitoring, aggregate token generation speed, processed token counts, and llama-swap Web UI access.
