PriorityQ — AI-Powered OPD Intelligence Platform

An intelligent outpatient department management system that uses Google Gemini AI to triage patients, prioritize queues dynamically, and generate clinical documentation in real-time.

---

Features

🏥 AI-Powered Patient Triage

• Patients describe symptoms via text or voice at check-in
• Gemini Flash analyses the symptoms and assigns a severity score (1–100)
• AI estimates the predicted consultation duration

📊 Live Priority Queue

• The doctor's dashboard shows a real-time queue sorted by Priority Score
• Priority Score formula: S + (T × 0.5) — Severity + Waiting Time (minutes)
• Implements an aging algorithm to prevent low-severity patients from waiting indefinitely
• Queue updates automatically every 60 seconds and instantly via WebSocket on any new check-in

🎙️ Voice-to-Text (Speech-to-Text)

• Patients can dictate symptoms verbally instead of typing
• Doctors can record consultations in real-time
• Audio is captured via the browser's MediaRecorder API and transcribed by Gemini AI on the backend

📋 AI Clinical Documentation (SOAP Notes)

• After a consultation, the doctor pastes or records the doctor-patient dialogue
• Gemini generates structured SOAP notes (Subjective, Objective, Assessment, Plan)
• AI also drafts a prescription with medication name, dosage, frequency, and duration
• A general advice and follow-up summary is also generated

🔌 Real-Time Updates

• Socket.IO pushes queue updates to all connected doctor dashboards simultaneously
• A live connection indicator shows socket status

---

Tech Stack

Layer	Technology
Frontend	React (Vite), Vanilla CSS, Lucide Icons, Socket.IO Client
Backend	Node.js, Express.js
Database	MongoDB (via Mongoose)
AI	Google Gemini Flash (@google/generative-ai)
Real-time	Socket.IO
Fonts	DM Sans, DM Serif Display, JetBrains Mono

---

Project Structure

PriorityQ/
├── backend/
│   ├── controllers/
│   │   ├── triageController.js      # Patient check-in & AI triage logic
│   │   ├── queueController.js       # Priority score calculation & broadcast
│   │   └── notesController.js       # SOAP note & prescription generation
│   ├── models/
│   │   └── PatientVisit.js          # Mongoose schema (severity, notes, prescription)
│   ├── routes/
│   │   └── api.js                   # All REST API endpoints + /transcribe
│   ├── services/
│   │   └── aiService.js             # Gemini AI: triage, SOAP notes, audio transcription
│   ├── server.js                    # Express + Socket.IO + MongoDB connection
│   ├── .env                         # Environment variables (API key, DB URI)
│   └── package.json
│
└── frontend/
    ├── src/
    │   ├── components/
    │   │   ├── PatientIntake.jsx     # Check-in form with voice input
    │   │   ├── DoctorDashboard.jsx  # Live priority queue for doctors
    │   │   └── ConsultationRoom.jsx # Dialogue input, SOAP notes, prescription
    │   ├── services/
    │   │   ├── api.js               # Axios API helpers
    │   │   └── socket.js            # Socket.IO client singleton
    │   ├── App.jsx                  # Router + layout + nav
    │   └── index.css                # Design system (glassmorphism, tokens)
    └── index.html

---

Getting Started

Prerequisites

• Node.js v18+
• MongoDB (running locally on mongodb://localhost:27017)
• A Google Gemini API key from Google AI Studio

⚠️ Important: Create your API key by clicking "Create API key in new project" in AI Studio. Do not use an existing Google Cloud project — it may have billing restrictions that block the free tier.

---

1. Clone & Install

# Install backend dependencies
cd backend
npm install

# Install frontend dependencies
cd ../frontend
npm install

2. Configure Environment

Create / update backend/.env:

PORT=5000
FRONTEND_URL=http://localhost:5173

MONGO_URI=mongodb://localhost:27017/opd_platform

GEMINI_API_KEY=your_api_key_here
AI_MODEL=gemini-2.0-flash

3. Start MongoDB

Make sure your local MongoDB service is running:

# Windows (run as Administrator)
net start MongoDB

4. Run the Application

Open two terminals:

# Terminal 1 — Backend
cd backend
npm run dev

# Terminal 2 — Frontend
cd frontend
npm run dev

• Frontend: http://localhost:5173
• Backend API: http://localhost:5000

---

How to Use

Patient Check-In (Patient Intake page)

1. Enter the patient's name and age
2. Describe symptoms by typing or clicking "Speak" to use voice
3. Click "Check In & Assess"
4. The AI returns a severity score (1–100), a label (e.g., Moderate), and an estimated consultation time

Doctor's Queue (Doctor Dashboard)

1. See all waiting patients sorted by Priority Score (highest first)
2. The #1 patient is highlighted as "Next"
3. The queue auto-refreshes every 60 seconds; new check-ins trigger an instant update via Socket.IO
4. Click "Consult" to open the consultation room for any patient

Consultation Room

1. Type or record the doctor-patient dialogue
2. Click "Generate Clinical Notes & Prescription"
3. The AI produces structured SOAP notes and a draft prescription
4. Click "Complete & Discharge" when done — the patient's status updates to COMPLETED

---

Priority Algorithm

Priority Score = Severity Score + (Minutes Waiting × 0.5)

Term	Description
S	AI-assigned severity (1–100). A crushing chest pain might score 92, a mild cold ~15
T	Minutes the patient has been waiting since check-in
× 0.5	Aging weight — ensures low-severity patients still rise over time (starvation prevention)

Example: A patient with severity 25 who has waited 40 minutes has a score of 25 + 20 = 45, outranking a new patient with severity 40 who scores 40 + 0 = 40.

---

API Reference

Method	Endpoint	Description
GET	/api/health	Health check
POST	/api/patients/checkin	Check in a patient + run AI triage
GET	/api/patients	List all patients
GET	/api/patients/:id	Get patient details
PUT	/api/patients/:id/status	Update patient status
GET	/api/queue	Get current sorted queue
POST	/api/patients/:id/notes	Generate SOAP notes + prescription
PUT	/api/patients/:id/complete	Mark patient as discharged
POST	/api/transcribe	Transcribe audio blob via Gemini

---

Known Limitations

• Voice transcription requires a working Gemini API key (free tier: ~50 req/day)
• The Web Speech API fallback is not used — all STT goes through Gemini on the backend
• Browser support: MediaRecorder works in Chrome, Edge, and Firefox
• MongoDB must be running locally; there is no cloud DB fallback

---

License

MIT — feel free to use, modify, and distribute.