[!WARNING]
Service status (9 Oct 2026): pumpfoil.org is wrongly blocked by the DNS4EU filtering resolvers.
If the app or the website shows “Hostname pumpfoil.org not verified” with a certificate for
*.joindns4.eu, your phone or network uses the filtering DNS from DNS4EU, which currently sends
pumpfoil.org to its warning page. Every other DNS service resolves it correctly. We have reported
the false positive (DNS4EU / Whalebone ticket 58102).
Workaround until it is fixed: on Android set Settings → Network → Private DNS to
Automatic (or to unfiltered.joindns4.eu); on other devices use your provider's DNS or the
unfiltered DNS4EU resolver 86.54.11.100. Updates on this page.
Pumpfoil
Record and analyze pump foiling sessions from your sports watch — GPS track, foiling distance, pump cadence and glide phases. Garmin, Wear OS, Apple Watch, Amazfit (Zepp OS) and Huawei (beta), with native iOS/Android apps and a web PWA.
pumpfoil.org · License: AGPL-3.0
Pump foiling means riding a hydrofoil with no wind, waves or motor — you stay up purely by
pumping the board rhythmically. Pumpfoil records these sessions on your watch and turns the
raw GPS + accelerometer data into a detailed analysis: how long you were actually foiling, how
efficiently you pumped, and where you rode.
The signature feature: your track is colored exactly over the foiling phases (by speed, heart
rate or pump cadence) with every detected pump stroke marked right on the route.
Features
- Watch recording on every platform — a Connect IQ app for all ~78 Garmin
devices (fēnix, Forerunner, epix, Instinct, vívoactive, …), plus native Wear OS,
Apple Watch and Amazfit (Zepp OS) recorder apps. They capture GPS + heart rate (Garmin
also raw acceleration) and upload the raw data.
- Native companion apps — full iOS and Android apps (sessions, map, track preview,
per-run stats, history with trend charts, community) alongside the installable web PWA.
- Automatic analysis — foiling phases, distance, pump cadence and glide phases per session,
detected server-side from GPS + accelerometer (ML model + GPS state machine).
- Color-coded track — foiling segments colored by speed / heart rate / pump frequency, with pump markers.
- FIT upload — analyze old activities too: drop in a
.fit file (or Garmin's ZIP export).
- Labeling UI — mark pump / glide / not-foiling ranges to build training data for the ML model.
- Community & history — compare sessions over time, share runs; web UI and apps in 18 languages.
- One-click watch download — the site detects your watch model and serves the matching build.
A look inside the app
| |
|---|
| Track your progress over time | Community records |
| |
| Community records per spot | Sessions — yours & everyone's |
| |
| Detailed per-run stats | |
| |
Apps & watches
Native apps on every platform — one account, one analysis backend.
📱 Android phone
📱 iOS phone
📱 No watch? Record with the phone (Android & iOS app — strapped to the board or in a pocket)
⌚ Wear OS
⌚ Apple Watch
⌚ Garmin (all ~78 Connect IQ devices)
⌚ Amazfit (Zepp OS)
⌚ Huawei (HarmonyOS — new, beta)
Architecture
| Directory | Stack | Purpose |
|---|
watch/ | Monkey C (Connect IQ) | Garmin recorder (all ~78 devices): records GPS + raw accelerometer, uploads raw data |
android/ | Kotlin · Jetpack Compose | One Gradle project: Android phone app (:app) + Wear OS recorder (:wear) |
watch-apple/ | Swift · SwiftUI | iOS companion app (Sources-iOS/) with embedded Apple Watch recorder (Sources/) |
watch-zepp/ | JavaScript · Zepp OS (@zeppos/zml) | Amazfit recorder: records GPS + heart rate, uploads via the Zepp app |
server/ | Python · FastAPI · PostgreSQL · numpy/scipy/scikit-learn | Ingest, immutable raw storage, foiling/pump detection, REST API |
web/ | React · Vite · TypeScript · Tailwind · Leaflet | Installable PWA: sessions, map, charts, labeling, community |
deploy/ | systemd · Apache | Service unit, reverse-proxy config, backup timers |
Raw session data is always stored complete and unmodified, so any future detection model can be
re-run on old sessions. Detection runs server-side (fast iteration in Python, no watch recompile).
Development
Server
cd server
python3 -m venv .venv && . .venv/bin/activate
pip install -e .
cp .env.example .env # then set DATABASE_URL (Postgres is required)
uvicorn app.main:app --reload --port 8090
# API docs: http://localhost:8090/api/docs (only with FOIL_API_DOCS=1)
Web
cd web
npm install
npm run dev # Vite dev server, proxies /api to :8090
Garmin watch — requires the Garmin Connect IQ SDK
and a developer key (see watch/setup-sdk.sh). The SDK is not included (distributed by Garmin
under its own license). Build one device with watch/build.sh <device>, or all manifest devices
with watch/build-all.sh (also generates the download catalog the website serves).
Android phone + Wear OS — one Gradle project under android/ (modules :app and
:wear), Kotlin + Jetpack Compose. Build/verify with ./gradlew :app:compileDebugKotlin /
:wear:compileDebugKotlin (JDK 21).
iOS app + Apple Watch — watch-apple/, Swift + SwiftUI, generated with
xcodegen (project.yml); built in Xcode on macOS.
The raw upload format (GPS + int16 accelerometer chunks) is specified in
docs/data-format.md.
Privacy
Pumpfoil is a community platform: by design, your sessions are visible to other users in the
community feed. What it will never do is sell your data or hand it to third parties for
advertising or tracking. This repository contains source code only — no user accounts,
recordings, databases or Garmin SDK files.
License
Pumpfoil is free software, licensed under the GNU Affero General Public License v3.0
(AGPL-3.0). In short: you may use, study, modify and redistribute it, but if you run a
modified version as a network service, you must make your source available to its users.
It is a community project — contributions are welcome.
