CareMeet – AI-Assisted Telemedicine Platform

Overview

CareMeet is a full-stack telemedicine platform that connects patients with verified doctors through video consultations, supported by an AI-powered symptom analysis system.

The platform handles the complete consultation lifecycle — from helping users understand their symptoms using AI, to booking appointments, managing payments through a credit system, and conducting video consultations.

This project focuses on building a real-world healthcare workflow with reliable backend handling. Instead of being just a CRUD application, it addresses practical challenges such as scheduling conflicts, safe payment handling, role-based access, and integration with external systems.

---

System Flow

User → AI Analysis → Doctor Selection → Booking → Payment → Video → Payout

Step-by-step Flow

1. Symptom Input

   • User enters symptoms through the AI assistant.

2. AI Processing

   • Backend sends structured prompt to Gemini API.

   • Response is parsed to extract:

     • urgency level
     • recommended medical specialties

3. Doctor Selection

   • User browses verified doctors filtered by specialty.
   • Available slots are dynamically generated.

4. Booking Validation

   • Backend checks:

     • user has sufficient credits
     • slot availability (overlap query)
     • minimum booking buffer (2 hours)

5. Atomic Booking Execution

   • A database transaction is executed:

     • deduct credits from patient
     • add credits to doctor
     • create appointment record

   • If any step fails → everything is rolled back

6. Video Consultation

   • Secure Vonage session is generated
   • Doctor and patient join at scheduled time

7. Payout Flow

   • Doctors accumulate credits
   • Admin processes payout requests safely with balance validation

---

Core Features

🤖 AI Symptom Analysis

• Uses Google Gemini to analyze user symptoms
• Returns structured output (urgency + specialty)
• Includes fallback handling for API failures

---

📅 Appointment Scheduling

• Doctors define availability
• System generates 30-minute slots dynamically
• Prevents overlapping bookings using database checks
• Enforces minimum 2-hour buffer before booking

---

💳 Credit-Based Payment System

• Users purchase credits via subscription
• Each consultation deducts fixed credits
• Credits are tracked using a ledger-style transaction model
• Doctor earnings are accumulated and later paid out

---

🎥 Video Consultation Integration

• Uses Vonage Video API (WebRTC)
• Session IDs and tokens generated securely on server
• Supports real-time video/audio communication

---

🔐 Role-Based Access Control

• Roles: PATIENT, DOCTOR, ADMIN
• Doctors require verification before becoming active
• Access control enforced in both server logic and UI

---

System Architecture

The application follows a modern full-stack architecture using Next.js:

• Next.js Server Actions

  • Handle data mutations securely on the server
  • Reduce need for separate API endpoints

• Prisma + PostgreSQL

  • Manages relational data (users, appointments, transactions)
  • Ensures type-safe database operations

• Clerk Authentication

  • Handles user authentication and session management
  • Supports role-based workflows

• External Integrations

  • Gemini API → symptom analysis
  • Vonage API → video sessions

---

Key Engineering Decisions

Atomic Transactions for Booking

Booking requires multiple updates (credits + appointment). These are executed inside a single database transaction to ensure:

• no credit loss
• no partial state in case of failure

---

Consistent Payout Handling

Payout processing validates and updates doctor balance inside a transaction, preventing:

• duplicate payouts
• inconsistent balances

---

Scheduling Conflict Prevention

Instead of relying on frontend checks, the system:

• queries overlapping time slots in database
• ensures no double booking under concurrent requests

---

Input Validation with Zod

All critical inputs (booking, availability, payouts) are validated before hitting the database to prevent invalid data.

---

Role-Based Access Enforcement

Authorization checks are implemented in:

• server actions
• layout-level logic

This ensures restricted access to sensitive operations.

---

Edge Cases & Reliability

• Prevents double booking using backend overlap checks
• Rolls back booking if any step fails
• Enforces 2-hour buffer for realistic scheduling
• Handles invalid inputs and malformed requests safely
• Includes fallback handling for AI failures

---

Tech Stack

Frontend

• Next.js 15 (App Router)
• React 19
• Tailwind CSS
• Radix UI

Backend

• Node.js
• Next.js Server Actions

Database

• PostgreSQL
• Prisma ORM

Integrations

• Clerk (Authentication)
• Google Gemini (AI)
• Vonage Video API (WebRTC)

---

Setup Instructions

git clone https://github.com/Shikhar1504/Doctors_Appointment_Platform.git
cd Doctor
npm install

Create .env.local:

DATABASE_URL=...

NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=...
CLERK_SECRET_KEY=...

GEMINI_API_KEY=...

NEXT_PUBLIC_VONAGE_APPLICATION_ID=...
VONAGE_PRIVATE_KEY=...

npx prisma generate
npx prisma db push
npm run dev

---

Important Routes

• /api/ai-health → processes symptom input and returns structured AI response
• booking & payout logic handled via server actions

---

What This Project Demonstrates

• Full-stack application development with real-world workflows
• Backend correctness using transactions and validation
• Integration of AI and real-time communication systems
• Handling scheduling and payment logic reliably
• Building role-based multi-user systems

---

Built by Shikhar