Turn your old workout notes into a coach — then log every set live.
Private, offline-first strength tracking — import, log, coach, and program your training, all in your browser. No account, no server; on-device by default.
Open the app » · Install as a PWA · Report a bug · Request a feature
Note
On-device by default. No account, no sign-up, no cloud — your data lives in your browser (IndexedDB) and works fully offline once installed. The one optional exception is AI parse: if you choose to add your own free API key, the note text you ask it to parse is sent to that provider — nothing else, and only when you use it.
RepLog is a workout tracker that does two things really well:
- Imports the workout notes you already have — screenshots, pasted text, even a screen-recording of your notes app — reads them with on-device OCR, cleans up the mess, and turns years of scattered logs into a structured training history.
- Logs your workouts live — start a session, search an exercise, and record set-by-set with previous-session hints, an auto rest timer, a plate calculator, and reusable routines.
On top of that history it builds progress charts, a trends view (PR timeline + training-block comparison), a strength coach that scores how balanced your lifts are, and a program engine that generates a research-based weekly plan and auto-progresses it from what you lift.
It is 100% client-side: a React + TypeScript app that runs in your browser, stores data on your device, and installs as a Progressive Web App on your phone or laptop. Free and open source.
- 📸 Import notes three ways — upload screenshots, paste plain text, or drop in a phone screen-recording. In-browser OCR extracts the text; a relevance filter automatically keeps the workout notes and drops everything unrelated.
- 🧹 Smart messy-notes parser — normalizes inconsistent exercise names and pulls out dates, bodyweight, the training split, and every set's reps and weight. A review step lets you confirm or fix anything before it's saved.
- 🏋️ Live workout logger — start or resume a session, search exercises, and log set-by-set with previous-session hints, warm-up sets, RPE, an automatic rest timer, and a built-in plate calculator.
- 🔁 Routines & templates — save reusable routines and load them in one tap, then finish to commit the session to history.
- 📈 Progress that means something — all-time bests, estimated 1RM, and per-exercise progression graphs with metric toggles (heaviest weight, est. 1RM, volume, reps) and selectable time ranges.
- 🗂️ Full history — per-session set history, bodyweight trend, training-block and split history, and a complete workout log.
- 🧠 Strength coach — a 0–100 strength-balance score with per-area ring gauges, plus recommendation cards that flag underdeveloped muscles by comparing your lifts against established strength-ratio standards (e.g. overhead press vs. bench, squat vs. bench).
- 📊 Weekly volume insights — sets per muscle over the last week vs. evidence-based MEV/MAV ranges, with "add N sets here" recommendations.
- 🗓️ Program engine — generate a weekly plan from your goal, experience, days/week, split, session length, and recovery (sleep/stress). It picks exercises by muscle, sets rep ranges + RIR, auto-progresses with double progression + daily undulating periodization, and lets you swap any lift for a similar one.
- 📅 Trends — a PR timeline (every weight + est-1RM record, newest first) and side-by-side training-block comparison.
- 💾 Backup & restore — export all your data to a JSON file and import it on another device.
- 🤖 Optional AI parse — for messy notes the built-in parser can't crack, bring your own free OpenRouter key (stored on your device) to extract workouts with an LLM. Fully opt-in; the deterministic parser stays the default.
- 📱 Web + native — an installable PWA (offline on iOS/Android/desktop) plus a native Expo app in
mobile/sharing the same engine. - 🔒 Private by default — no account, no sign-up, no backend. Everything runs and stays on your device; the only optional network feature is AI parse, with your own key.
| Dashboard | Coach |
|---|---|
 |
 |
| Live logger | Import |
|---|---|
 |
 |
| Settings | Per-exercise progress |
|---|---|
 |
 |
Most people already have a training history — it's just trapped in notes apps, photos, and spreadsheets in a hundred different formats. RepLog's whole premise is that your past data is the most valuable input a tracker can have, so it meets you where that data already lives.
Import → parse → review → save. You hand RepLog a screenshot, some pasted text, or a screen-recording of your notes. It runs OCR in your browser, filters out anything that isn't a workout, and parses the rest into structured sessions — normalizing exercise names and extracting dates, bodyweight, splits, and sets/reps/weight. You get a review screen to fix anything before it's committed.
Log live, learn over time. Going forward you log workouts set-by-set in the app. Each new entry feeds the progress charts and the coach, so the picture of your training keeps sharpening with every session.
Coach by the numbers. The coach compares your key lifts against well-established strength-ratio standards and surfaces where you're out of balance — turning raw logs into a clear "here's what to work on" without any guesswork or guru opinions.
Program your training. Beyond tracking, RepLog builds a weekly program from a few inputs — using research-based volume landmarks (MEV/MAV), goal-appropriate rep ranges, and your recovery (sleep/stress) — then progresses it automatically from what you actually lift, and swaps exercises on request. Optionally refine the plan with your own AI key.
This is the part that matters most, so it's worth being explicit:
- On-device by default. OCR, parsing, charts, the coach, trends, and the program engine all run on-device — there's no backend to send your data to. The sole exception is the optional AI parse, which only runs when you enable it with your own key.
- No account, no sign-up. Open the app and start. The built-in parser, logger, coach, trends, and program engine need no key; AI parse is the one opt-in that does.
- Local storage. All your workouts live in your browser's IndexedDB. Clearing site data or uninstalling removes them; nothing lingers on a server.
- Works offline. Once installed as a PWA, RepLog loads and functions without a connection.
Open https://benasbarciauskas.github.io/RepLog/ in any modern browser. That's it — no install required to start logging.
Install it as a PWA so it works offline and lives on your home screen / dock:
- iPhone / iPad (Safari): tap Share → Add to Home Screen.
- Android (Chrome): tap the ⋮ menu → Install app / Add to Home screen.
- Desktop (Chrome / Edge): click the install icon in the address bar, or the ⋮ menu → Install RepLog.
Prerequisites: Node.js ≥ 20
git clone https://github.com/benasbarciauskas/RepLog.git
cd RepLog
npm install
npm run dev # open http://localhost:5173RepLog compiles to a folder of static files — host it on any static host or your own server, no backend required.
npm run build # outputs static assets to /dist
npm run preview # optional: preview the production build locallyThen deploy the contents of /dist to any static host (object storage + CDN, a static-site host, GitHub Pages, or a plain web server). Because there's no server-side component, there's nothing else to provision.
A native iOS/Android build lives in mobile/ — Expo + expo-router, sharing this repo's parser, analytics, and coach logic, with on-device expo-sqlite storage.
cd mobile
npm install
npx expo start # scan the QR code with Expo Go on your phoneSee TUTORIAL.md → Native app via Expo for the Expo Go workflow and building/submitting a real binary with EAS (expo.dev).
| Layer | Tools |
|---|---|
| UI | React 19, TypeScript |
| Build | Vite 7, vite-plugin-pwa |
| Styling | Tailwind CSS v4, shadcn/ui |
| Charts | Recharts |
| Local data | Dexie over IndexedDB (web), expo-sqlite (native) |
| OCR | Tesseract.js |
| Native app | Expo + expo-router, NativeWind |
| Delivery | Installable PWA (static site) + native iOS/Android via Expo |
Everything runs client-side. The OCR engine, the parser, the charts, and the coach all execute in the browser.
- Native mobile app (Expo / React Native) — core screens shipped in
mobile/, reusing the same on-device import, parsing, and coaching logic - Local export / import (JSON) for moving data between devices — on web and mobile; encrypted sync is next
- Optional cross-device sync (end-to-end encrypted)
- More import formats and smarter exercise-name normalization
- Program engine — research-based generator + auto-progression (double progression / DUP) + exercise swap + goal-based weekly volume insights (MEV/MAV)
- Trends — PR timelines + per-block comparisons
- Additional coach strength-ratio standards
See the open issues for the current backlog and to suggest something new.
Contributions are very welcome — bug reports, feature ideas, and pull requests all help.
- Fork the repo and create a branch:
git checkout -b feat/your-idea - Make your change and run the app locally (
npm run dev) - Commit and open a pull request describing what and why
Browse the open issues for good places to start, and see CONTRIBUTING.md for the workflow and conventions.
Distributed under the MIT License. See LICENSE for details.