PREHAB — AI-Powered Injury Prevention for Female Athletes

An autonomous AI agent that adapts training plans to the menstrual cycle, workload, and body signals — keeping athletes healthy and coaches informed.

Built for HackEurope 2026 · Live Demo →

---

The Problem

Female athletes are 2–6× more likely to suffer ACL injuries during the ovulatory phase of their menstrual cycle (Hewett et al., 2007). Yet most training programmes ignore hormonal variation entirely, applying the same load regardless of cycle phase, soreness, or recovery status.

What PREHAB Does

PREHAB is a cycle-aware training optimisation platform with an autonomous AI agent that:

1. Predicts injury risk — ML model scores each athlete 0–100 using training load, soreness, RPE, and menstrual cycle phase
2. Adapts training plans — automatically adjusts session intensity, type, and duration based on risk level and cycle phase multipliers
3. Escalates to coaches — flags high-risk athletes with evidence-backed explanations; coaches accept, modify, or reject changes
4. Learns from feedback — tracks coach acceptance rate and shifts between Full Autonomy, Dampened, and Suggest-Only modes
5. Integrates wearable data — syncs cycle tracking and activity metrics from Samsung Health / Galaxy Watch

---

Features

Athlete Dashboard

• Daily injury risk score with plain-language AI explanation
• Cycle-phase-aware training plan (original vs. AI-adjusted toggle)
• One-tap session logging with effort and muscle soreness tracking
• Samsung Health integration (steps, active minutes, calories)
• AI-generated adaptive goals

Coach Dashboard

• Squad-level risk snapshot (Critical / High / Medium / Low tiers)
• Real-time escalation alerts with accept/modify/reject workflow
• Per-athlete plan management with coach override audit trail
• Squad risk trend analytics over time
• Adaptive agent autonomy controls per athlete

AI Agent Loop

OBSERVE → PREDICT → ADJUST → ESCALATE → REFLECT

Step	What happens
Observe	Fetch training sessions, soreness, cycle phase, wearable data
Predict	ML model scores injury risk; identifies top contributing factors
Adjust	Modify weekly plan — swap session types, reduce intensity, add rest
Escalate	Critical risk triggers immediate coach alert with reasoning
Reflect	Evaluate coach feedback history; update agent policy mode

---

Tech Stack

Layer	Technology
Frontend	React 18 · TypeScript · Vite
UI	shadcn/ui · Tailwind CSS · Framer Motion
Backend	Supabase (PostgreSQL · Auth · Edge Functions)
AI/ML	XGBoost (risk scoring) · Google Gemini (plan explanations)
Wearables	Samsung Health simulation API
Billing	Paid.ai usage-based signals
Deployment	Lovable (frontend) · Supabase (backend)

---

Architecture

Athlete App  ──►  Supabase DB  ◄──  agent-runner (Edge Function)
                                          │
                              ┌───────────┴───────────┐
                              ▼                       ▼
                        risk_predictions        weekly_plans
                        risk_reports            agent_runs
                        escalations             feedback_events
                              │
                              ▼
                        paid-signal (Edge Function)
                              │
                              ▼
                          Paid.ai API

Edge Functions

Function	Purpose
agent-runner	Main agentic loop — risk scoring, plan adjustment, escalation, billing signals
paid-signal	Validates and forwards usage economics to Paid.ai
seed-demo-data	Seeds realistic demo athlete data for presentations

---

Getting Started

Prerequisites

• Node.js 18+
• A Supabase project (or use Lovable Cloud)

Installation

git clone https://github.com/gracesmith6504/prehabFrontend.git
cd prehabFrontend
npm install

Environment Variables

Copy the example environment file and fill in your values:

cp .env.example .env

Variable	Description
VITE_SUPABASE_URL	Your Supabase project URL
VITE_SUPABASE_PUBLISHABLE_KEY	Supabase anon/public key
VITE_SUPABASE_PROJECT_ID	Supabase project ID

Edge Function secrets (set via Supabase dashboard):

Secret	Description
PAID_API_KEY	Paid.ai API key for usage billing
GEMINI_API_KEY	Google Gemini for plan explanations
ML_ENDPOINT_URL	ML model endpoint for risk prediction

Run Locally

npm run dev

The app runs at http://localhost:8080.

---

Pages & Routes

Route	Description
/	Landing page — hero, features, pricing, FAQ
/auth	Login / signup with role selection (athlete or coach)
/dashboard	Athlete dashboard — risk gauge, today's session, goals
/plan	Weekly training plan with original vs. AI-adjusted toggle
/risk-report	Full risk report with ML explanation and research citations
/cycle-setup	Cycle configuration + Samsung Health sync
/coach	Coach command centre — squad risk, escalations, analytics
/coach/escalations	Active escalations with accept/modify/reject workflow
/coach/plans	Per-athlete plan templates and management
/coach/analytics	Squad risk trends and agent performance
/coach/settings	Demo data seeding for presentations

---

Project Structure

src/
├── components/          # Reusable UI components
│   ├── coach/           # Coach-specific components
│   └── ui/              # shadcn/ui primitives
├── hooks/               # Custom React hooks (auth, mobile, toast)
├── integrations/        # Supabase client and auto-generated types
├── lib/                 # Utilities and risk calculation engine
├── pages/               # Route-level page components
│   └── coach/           # Coach dashboard pages
└── assets/              # Images and static assets

supabase/
└── functions/           # Supabase Edge Functions
    ├── agent-runner/    # Main AI agent loop
    ├── paid-signal/     # Paid.ai billing integration
    └── seed-demo-data/  # Demo data seeder

public/                  # Static files served at root

---

Acknowledgements

• Research basis: Hewett et al. (2007), Gabbett (2016), Soligard et al. (2017)
• Built with Lovable, Supabase, shadcn/ui
• Agent economics powered by Paid.ai

---

License

This project was built for HackEurope 2026. All rights reserved.