How I built it
How I built Ruxi: a headless fork of Rufus’s C engine
Reinstalling Windows still costs €80-100 at a shop. Rufus is reliable but speaks GPT, UEFI and TPM. Ruxi wraps the same engine with a 9-screen wizard.
The problem was not Rufus
Rufus works. The issue is it assumes you know what a GPT partition table is, whether your board uses UEFI and why TPM might block the install. For most people that is not vocabulary: it is noise.
The decision was not to rewrite formatting from scratch. Rufus has been battle-tested on millions of USB drives. What was missing was a layer that turns technical choices into plain-language questions.
Architecture: Electron UI, C engine
Ruxi splits two worlds. The Electron wizard (React + TypeScript) guides the user screen by screen: pick ISO, select USB, confirm partition, review bypass options.
The engine is a fork of Rufus’s C code built in headless mode. No window: it reads JSON from stdin and streams progress and results on stdout. The UI only shows bars and human messages.
// Simplified headless engine contract
{
"action": "format",
"device": "\\\\.\\PhysicalDrive2",
"scheme": "GPT",
"filesystem": "NTFS",
"bypass_tpm": true,
"bypass_secure_boot": true
}
→ stdout: {"progress":64,"speed_mbps":18,"phase":"writing"}Details that matter
System disk lock: the engine filters the volume running Windows before listing devices. The UI asks for a typed confirmation with the disk name.
Documented TPM and Secure Boot bypass: not a hidden trick; the wizard explains each option and when you need it.
GPL-3.0 respected: the C engine fork keeps the original license. The installer was published on VirusTotal before the public release.
What I learned
80% of the value is translation, not the algorithm. The engine already existed; the product is the 9-screen flow, 7 languages and the messages when something fails.
Headless + JSON is the cheapest bridge between a legacy C engine and a modern UI. You do not need to rewrite 50,000 lines to ship.