Sense safety before you think it.
Track your body's pre-conscious sense of safety vs danger and recover before overload.
SafeSignal is a speculative near-future wellness platform that tracks neuroception — the body's pre-conscious sense of safety vs danger — and helps users detect unsafe nervous-system states earlier, intervening with subtle environmental adjustments before panic, dissociation, or shutdown.
graph LR
A[React Frontend] -->|API Proxy| B[Express Server]
B --> C[SQLite via Prisma]
B --> D[Gemini AI]
A -->|Web Bluetooth| E[Wearable Device]
Wearable / Simulation ──▶ Sensor Reading ──▶ Neuroception Engine ──▶ State (Safe/Guarded/Overloaded)
│
▼
Interventions + AI Insights
│
▼
SQLite DB (per user)
| Component | Files | Purpose |
|---|---|---|
| Auth | auth.service.ts, auth.routes.ts |
JWT + bcrypt, register/login |
| Neuroception | neuroception.service.ts, session.routes.ts |
6-sensor scoring engine, sessions |
| AI | gemini.service.ts, insight.routes.ts |
Gemini insights with fallback |
| Middleware | auth.middleware.ts, error.middleware.ts |
JWT verification, error handling |
| Database | schema.prisma |
User, Session, SensorReading, Intervention, DailyReflection |
| Page | Key Features |
|---|---|
| LandingPage | Animated halo, use case tabs, safeguards |
| AuthPage | Login/signup with JWT |
| DashboardPage | Live sensors, body halo, wearable, interventions |
| ReflectionPage | Timeline, AI insights |
| ProfilePage | Settings, safeguards |
| Component | Purpose |
|---|---|
| WearablePicker | Branded device selector modal |
| useWearable | Web Bluetooth API hook |
| Layout | Responsive navbar + mobile drawer |
- Monitors 6 micro-signals: breath pace, jaw tension, posture collapse, skin conductance, voice strain, motion restlessness
- Weighted scoring algorithm computes nervous system state with confidence percentage
- 3-state model: Safe → Guarded → Overloaded with animated body-map halo visualization
- Branded device picker supporting popular wearables:
- Apple Watch, Fitbit, Garmin, Samsung Watch, Mi Band/Amazfit
- Plus health apps: Google Fit, Fitbit App, Apple Health, Samsung Health
- Web Bluetooth API connects to BLE heart rate monitors
- Derives neuroception signals from real heart rate + HRV data
- Automatic fallback to simulated data when no device is connected
- Grounding haptic pulses from wearable
- Quiet route suggestions (less crowded paths)
- Non-essential alert suppression
- 60-second micro-reset with guided breathing
- Window-of-tolerance timeline showing which moments drained or restored you
- Gemini AI-powered personalized insights, recommendations, and encouragement
- Activity breakdown with state tracking per time block
| Safeguard | Description |
|---|---|
| Consent-first sensing | No passive monitoring of others |
| Private by default | All data encrypted on-device |
| No truth claims | Confidence-based language only |
| Manual override | Mute all interventions instantly |
| Emergency mode | Grounding + trusted-contact alerts |
| Anti-misuse | No employer/school/partner access |
- Glassmorphism UI with dark/light mode
- Mobile-responsive with hamburger menu and slide-in drawer
- Sound effects (click, send, received) with toggle
- Smooth animations and micro-interactions throughout
| Layer | Technology |
|---|---|
| Frontend | React 19, TypeScript, Vite, Zustand, React Query, CSS Modules |
| Backend | Node.js, Express, Prisma ORM, SQLite |
| AI | Google Gemini API (with intelligent fallback) |
| Hardware | Web Bluetooth API (BLE Heart Rate Profile) |
| Security | Helmet, CORS, Rate Limiting, Compression, bcrypt, JWT |
- Node.js 18+
- npm
# 1. Clone the repository
git clone https://github.com/ritikiitg/SafeSignal.git
cd SafeSignal
# 2. Setup the server
cd server
npm install
cp .env.example .env # Edit with your own JWT_SECRET
npx prisma db push
# 3. Setup the client
cd ../client
npm install# Terminal 1 — Start the API server
cd server
npm run dev # Runs on http://localhost:3001
# Terminal 2 — Start the frontend
cd client
npm run dev # Runs on http://localhost:5173Open http://localhost:5173 in your browser.
SafeSignal/
├── client/ # React 19 + TypeScript + Vite
│ ├── public/assets/ # Logos, sound effects
│ │ ├── logo_favi.png # App favicon/nav logo
│ │ ├── click.wav # UI click sound
│ │ ├── send.wav # Send action sound
│ │ └── received.wav # State change sound
│ └── src/
│ ├── components/
│ │ ├── layout/ # Layout.tsx — Navbar, mobile drawer
│ │ └── wearable/ # WearablePicker.tsx — Device selector
│ ├── hooks/
│ │ ├── useSound.ts # Audio feedback hook
│ │ └── useWearable.ts # Web Bluetooth integration
│ ├── pages/
│ │ ├── LandingPage.tsx # Hero, use cases, safeguards
│ │ ├── AuthPage.tsx # Login / Register
│ │ ├── DashboardPage.tsx # Live detection + interventions
│ │ ├── ReflectionPage.tsx # Daily timeline + AI insights
│ │ └── ProfilePage.tsx # Settings + preferences
│ ├── services/api.ts # API client with JWT
│ ├── stores/ # Zustand (auth + neuroception)
│ └── index.css # Design system + glassmorphism
│
├── server/ # Express + TypeScript + Prisma
│ ├── prisma/schema.prisma # Database models
│ └── src/
│ ├── routes/ # auth, user, sessions, insights
│ ├── services/
│ │ ├── auth.service.ts # JWT + bcrypt authentication
│ │ ├── neuroception.service.ts # State engine + scoring
│ │ └── gemini.service.ts # AI insights with fallback
│ ├── middleware/ # Auth + error handling
│ └── index.ts # Express app entry
│
├── idea.txt # Product concept
├── walkthrough.md # Build walkthrough & verification
└── README.md
SafeSignal uses a weighted scoring algorithm to compute nervous-system state from 6 body signals:
| Signal | Weight | Source |
|---|---|---|
| Breath pace | 20% | Wearable HRV / Simulated |
| Skin conductance | 20% | Wearable HR / Simulated |
| Jaw tension | 18% | Camera / Estimated |
| Posture collapse | 15% | IMU / Estimated |
| Voice strain | 15% | Microphone / Estimated |
| Motion restlessness | 12% | Wearable accelerometer / Simulated |
State thresholds:
- Safe (score 0–30): Body is open and regulated
- Guarded (score 30–65): Early stress signals detected
- Overloaded (score 65+): System approaching or in overload
| Variable | Description | Default |
|---|---|---|
PORT |
Server port | 3001 |
DATABASE_URL |
SQLite database path | file:./dev.db |
JWT_SECRET |
JWT signing secret | (required) |
JWT_EXPIRES_IN |
Token expiration | 7d |
GEMINI_API_KEY |
Google Gemini API key | (optional) |
Ritika enters a packed train. SafeSignal detects rising threat response before she notices. Wearable sends a grounding pulse; app suggests a calmer exit.
Outcome: Overload prevented before panic peaks.
During a meeting, voice flattens and posture tightens. SafeSignal drops from Safe to Guarded. It delays notifications and activates breath pacing.
Outcome: She stays present instead of shutting down.
Ritika reviews her daily window-of-tolerance timeline. She sees which moments drained her and which restored her. The app recommends behavior changes.
Outcome: Stronger self-understanding and better planning.
| Check | Result |
|---|---|
Client vite build |
✅ 109 modules, 0 errors, 1.30s |
| Server runs | ✅ Port 3001, Helmet + Rate Limiting + Compression |
| Client dev server | ✅ Vite 7.3.1, HMR active |
| npm install (client) | ✅ 76 packages, 0 vulnerabilities |
| npm install (server) | ✅ 278 packages |
| Prisma DB push | ✅ SQLite database ready |
- Real-time camera-based jaw tension detection
- Native mobile app (React Native)
- Apple HealthKit & Google Health Connect server-side API integration
- Group session mode for therapists
- Export data to PDF for clinical use
- Multi-language support
Built for the Devpost Hackathon 2026
Powered by Google Gemini · Built with React + Express + Prisma