Phase 2 of the tPlan research project: a multi-user web application exploring AI-driven workout generation from training methodology books.
This app is part of a lab experiment investigating whether AI agents can read training books and produce correct, progressive workout programs. See the main README for the full research framing.
- Generates workouts from parsed training methodology (Convict Conditioning, Dumbbell Gymnastics)
- Tracks progress with difficulty ratings, reps, notes, and body weight
- Auto-progresses exercises based on the source book's transition rules
- Supports 4 languages (English, Russian, Latvian, Spanish)
- Social features — challenges and leaderboards for motivation research
| Layer | Technology |
|---|---|
| Frontend | React + TypeScript + Vite |
| Backend | Azure Functions (Node.js) |
| Database | Azure Cosmos DB |
| Hosting | Azure Static Web Apps |
| Auth | Google OAuth via SWA |
npm install
npm run devDeployed to Azure SWA: tplan.naurolabs.com
The planned next step is allowing users to upload their own training books (PDF/markdown) and have the system extract a structured program — exercises, levels, progression rules — that others can follow.