bun-monitor

AIStation Monitor — Bun-based local LLM and AI Chat system telemetry and model dashboard

bun-monitor is a high-performance, lightweight telemetry dashboard designed for the modern AI workstation. Built with the speed of the Bun runtime and TypeScript, it provides a unified "command center" view of your local LLM environment—streaming real-time NVIDIA GPU metrics, hardware thermals, and Ollama model status directly to a sleek Tailwind-powered UI. Whether you're monitoring a heavy inference load or tracking system health across a Tailscale connection, this tool ensures your Linux AI stack stays cool and responsive.

Features

• Real-time NVIDIA GPU stats (fan, temp, power, memory, utilization)
• System thermals (CPU, SSD, VRM, pump, fans) via sensors
• Disk usage and load average
• Ollama model status (via Docker)
• Modern web UI (Tailwind CSS)
• Fast Bun TypeScript backend
• Docker-ready for easy deployment

Quick Start

Install dependencies:

bun install

Run the server:

bun index.ts

Open http://localhost:4000 in your browser.

Docker Usage

Build and run with Docker Compose:

docker compose up -d --build

View logs:

docker compose logs -f

Run sensors inside the container:

docker exec -it bun-monitor bash
sensors

Development

• Source code is mounted into the container for live development
• Edit TypeScript and HTML/CSS, refresh browser to see changes
• Console logs are visible via Docker logs

API

• /api/stats — Combined telemetry (GPU, system, disk, models)
• /api/fresh_stats — Uncached fresh telemetry

Requirements

• Bun runtime (see Bun)
• Docker (for containerized deployment)
• NVIDIA drivers and Docker GPU toolkit (for GPU stats)
• lm-sensors (for system thermals)

License

MIT License