GhostLap

The problem

A RaceBox gives you a CSV with tens of thousands of GPS and IMU samples per session. What a rider actually wants to know is simpler: where am I losing time, and what changed when I found it? The vendor’s own app answers some of that. It doesn’t let you line up laps from different days or bikes, add notes to a corner, or see your data next to your friends’.

What I built

GhostLap rebuilds every session from the raw export. It re-times laps against the track’s start/finish line, coming within 0.7 ms of RaceBox’s own timing. It then lines laps up by distance, not time, so any two laps can be compared corner by corner.

• Analysis workspace: racing lines on satellite imagery, linked speed, brake, cornering-g and lean charts, sector splits and a live delta.
• Lean angle from the IMU, with gravity correction, so the number means something on a motorcycle.
• Garage: bikes, tyres and a maintenance log driven by the time you’ve actually ridden.
• Social layer: a shared lap database, teams, rider profiles, shareable analyses and weekend report cards.
• DuelVision and Studio: two laps raced head-to-head on the map, and telemetry overlays rendered onto your on-board video, entirely in the browser.
• A companion mobile app that syncs sessions straight from the device over Bluetooth.

Under the hood

It’s a TypeScript monorepo:

• A dependency-free core package handles parsing, lap detection, alignment, lean and reports. The API, the web app and the mobile app all share it.
• The API is Fastify 5 on Node’s built-in SQLite.
• The web app is React with MapLibre and uPlot, which keeps 25 Hz charts smooth.

It runs on a single small VPS behind Caddy, with nightly snapshots pulled back to the homelab NAS.

Outcome

GhostLap is live at ghostlap.app, and I use it after every track day. The demo here runs the real frontend on real sessions from my own bike. The API is replaced by an in-browser copy, so you can poke at everything without an account.