nfolens

A lightweight terminal UI for real-time Linux system stats, built with FTXUI. Reads directly from /proc — no external dependencies for data collection.

Features

• Hostname — read from /etc/hostname
• CPU usage — computed from /proc/stat deltas, plus clock speed from /proc/cpuinfo
• Memory usage — total and used, from /proc/meminfo
• Network throughput — combined upload/download rate across eth*, wl*, and en* interfaces, from /proc/net/dev
• Uptime — from /proc/uptime
• Live-updating gauges with a green → yellow → red gradient based on load
• Runs on a background polling thread; the UI thread only re-renders on change

Requirements

• Linux (relies on /proc and /etc/hostname)
• FTXUI
• CMake (recommended for pulling in FTXUI)

Building

Using CMake with FTXUI fetched as a dependency:

mkdir build
cmake -S . -B ./build -DCMAKE_BUILD_TYPE=Release
cmake --build ./build

If you choose to build and link FTXUI yourself, ftxui-component must be first in the linking order relative to the other FTXUI libraries, i.e.

g++ ./src/main.cpp -O2 -DNDEBUG -lftxui-component -lftxui-dom -lftxui-screen -lpthread -o ./build/main

Usage

./build/main [refresh_ms]

• refresh_ms — polling interval in milliseconds (default: 1000)
• Press q to quit

Example, refreshing every 500 ms:

./main 500

How it works

nfolens runs two threads:

1. Worker thread — polls /proc/uptime, /proc/cpuinfo, /proc/meminfo, /proc/net/dev, and /proc/stat on each interval, computes deltas (CPU busy time, network bytes/sec), and writes the results into a mutex-guarded shared struct.
2. UI thread — an FTXUI render loop that redraws whenever the worker thread posts a "changed" event, reading a locked snapshot of the shared state.

CPU usage is derived from the change in total vs. idle jiffies between two /proc/stat reads, rather than a single instantaneous read. Network rate is derived similarly from cumulative byte counters in /proc/net/dev.

Notes / Limitations

• Currently Linux-only (depends on procfs).
• Network interfaces are matched by name prefix (eth, wl, en); interfaces named differently (e.g. custom VPN/tunnel names) won't be counted.
• Assumes a single CPU frequency line format from /proc/cpuinfo; behavior only tested on x86_64 architecture.