serverpeek 1.3.0b1

Live-updating web dashboard for server monitoring

serverpeek

A live-updating web dashboard for server monitoring — CPU, memory, processes, Docker containers, and network ports in your browser.

Why?

Checking server health usually means SSH-ing in and running htop, docker ps, ss -tlnp, and half a dozen other commands. serverpeek gives you all of that in a single, auto-refreshing browser dashboard that you can leave open on a second monitor.

Features

• Machine & OS info — hostname, platform, architecture, CPU model, uptime
• CPU usage — overall with colour-coded bar (green → yellow → red), per-core bars on Linux
• Memory usage — physical and swap with macOS breakdown (app/wired/compressed), excludes file cache
• Load average — 1, 5, and 15 minute load averages
• Top processes — sorted by combined CPU + memory usage, grouped by parent process
• Smart process names — identifies scripts behind interpreters (python, node, ruby, perl, java) with runtime tags
• Docker containers — shows internal container processes via docker top, merged into unified process view
• Listening ports — non-localhost network ports with associated process names
• CPU & memory history graphs — 2 minute rolling history, shared across all viewers
• Live updates — Server-Sent Events push new data every 2 seconds
• Efficient — single background collector, sleeps when no clients are connected
• Beautiful dark theme — clean, modern kiosk-friendly dashboard (fits in one screen, no scrolling)
• Zero config — just run it and open the URL
• Single binary — statically-linked Go executable, no runtime dependencies

Quick Start

Install

pip install serverpeek

Or run directly with uv:

uvx serverpeek

Run

serverpeek

Then open http://localhost:8080 in your browser.

Options

serverpeek --port 9090        # Custom port
serverpeek --host 127.0.0.1   # Bind to localhost only
serverpeek --help             # Show all options

How It Works

serverpeek starts an HTTP server that serves a single-page dashboard. A background goroutine collects system snapshots every 2 seconds and shares them with all connected clients via Server-Sent Events (SSE). When no clients are connected, the collector sleeps. New clients receive the full 2-minute history buffer so graphs are populated immediately.

System information is gathered using OS-native interfaces (/proc on Linux, sysctl/vm_stat on macOS) and subprocess calls for Docker and network port information. On macOS, memory usage excludes file cache (reports active + wired + compressed instead).

The dashboard is a self-contained HTML page with embedded CSS and JavaScript — compiled into the binary via go:embed. No build tools, no npm, no bundlers. Designed for kiosk use: everything fits in a single non-scrolling screen.

Licence

Released under the Unlicense — public domain.