Real-time peripheral IV infiltration surveillance system. Streams physiological sensor data from an ESP32 to a Node.js backend and displays it on a React dashboard with live trend charts and clinical alert interpretation.

ESP32 (FSR + DS18B20)
       │  POST /data (JSON, every 1.5 s)
       ▼
Node.js + Express + Socket.io  (port 3001)
       │  WebSocket broadcast
       ▼
React Dashboard  (port 5173)

Sensors:

Risk scoring (0–100):

risk_score = min(temp_drop × 0.6 × 5, 60) + min(pressure_rise × 0.4 × 0.1, 40)

• Node.js ≥ 18
• npm ≥ 9
• Arduino IDE 2.x (for ESP32 firmware, optional)

cd backend
cp .env.example .env   # adjust PORT / FRONTEND_URL if needed
npm install
npm run dev            # starts on http://localhost:3001

The server automatically seeds three simulated patients:

• patient_001 — stable, no infiltration
• patient_002 — gradual developing infiltration (risk rises over ~8 minutes)
• patient_003 — established elevated risk

cd frontend
npm install
npm run dev            # opens at http://localhost:5173

{
  "device_id":     "patient_001",
  "temperature":   36.4,
  "pressure":      812,
  "temp_drop":     1.1,
  "pressure_rise": 58,
  "risk_score":    72,
  "alert_level":   "elevated",
  "timestamp":     1711039200
}

Only device_id, temperature, and pressure are required.

DS18B20 DATA  →  GPIO 4  (with 4.7 kΩ pull-up to 3.3 V)
FSR signal    →  GPIO 34 (voltage divider: FSR + 10 kΩ to GND)

1. Open esp32/iv_monitor.ino in Arduino IDE.
2. Install libraries via Library Manager:
   • OneWire (Paul Stoffregen)
   • DallasTemperature (Miles Burton)
   • ArduinoJson (Benoit Blanchon)
3. Edit the configuration block at the top of the sketch:

const char* WIFI_SSID  = "YOUR_WIFI_SSID";
const char* WIFI_PASS  = "YOUR_WIFI_PASSWORD";
const char* SERVER_URL = "http://192.168.1.x:3001/data";  // your machine's LAN IP
const char* DEVICE_ID  = "patient_001";

4. Select board: ESP32 Dev Module → Upload.
5. Open Serial Monitor at 115200 baud to watch calibration and POST status.

The device calibrates over 15 samples (~5 s) before transmitting. Keep the sensor undisturbed during calibration.

hackduke-2026/
├── backend/
│   ├── server.js               # Express + Socket.io entry point
│   ├── routes/
│   │   ├── data.js             # POST /data
│   │   └── patients.js         # GET /patients endpoints
│   ├── db/
│   │   ├── store.js            # In-memory time-series store
│   │   └── mockData.js         # Demo data generator (3 patients)
│   ├── middleware/
│   │   └── validate.js         # Input validation
│   └── .env.example
├── frontend/
│   ├── src/
│   │   ├── App.jsx             # Root layout, socket + REST wiring
│   │   ├── components/
│   │   │   ├── PatientList.jsx
│   │   │   ├── PatientCard.jsx
│   │   │   ├── LiveMonitor.jsx  # Main monitoring panel
│   │   │   ├── AlertPanel.jsx   # Clinical interpretation
│   │   │   ├── MetricsRow.jsx   # Temp / pressure / risk gauge
│   │   │   └── TrendChart.jsx   # Recharts line chart with smoothing
│   │   └── utils/
│   │       └── formatters.js    # Alert config, messages, time formatting
│   └── tailwind.config.js
└── esp32/
    └── iv_monitor.ino

The store interface (upsertReading, getPatient, getAllPatients, getHistory) is intentionally minimal. To swap in Firestore:

1. npm install firebase-admin in /backend
2. Create backend/db/firestore.js implementing the same four exports
3. Update the require in routes/data.js and routes/patients.js

• Charts apply a 3-point rolling average to reduce FSR noise while preserving real trends.
• patient_002 naturally progresses Normal → Monitor → Elevated → Critical over ~8 minutes — useful for live demos.
• Alert messages integrate both temperature and pressure into a single sentence; no raw algorithm logic is exposed.
• Critical status indicators use CSS animate-pulse for immediate visual salience.