CareSlot

Clinic Appointment Booking Agent
Team demo | PS3 | SRM Agentic AI Hackathon

The problem

Patients of a multi-speciality clinic call the front desk just to book, move or cancel appointments. Lines stay busy and desk staff cannot keep up.

What our app does

• Lets patient Arjun Mehta (P001) ask for care in plain English
• Suggests the right department from symptoms (never diagnoses or suggests medicine)
• Shows free slots and books, reschedules or cancels appointments
• Marks a slot booked exactly once; cancelling frees it again
• Includes a stretch doctor day view for the signed-in doctor’s schedule
• Password auth for patients and doctors, then issues a JWT with role + userId
• Clinic console UI: Dashboard, Knowledge Center, and Reports alongside Booking

Live demo

• Vercel link: https://careslot-agent.vercel.app
• Landing: / with Sign in / Sign up
• Register: /register creates a new patient or doctor (password + confirm password)
• Login: /login with id + password (demo: P001 / DR01, password demo1234)
• Demo prompts:
  • I have a skin rash, book me with a doctor on Friday morning.
  • Move my appointment to the next free slot.
  • Which Tamil-speaking doctors are free tomorrow?
  • What medicine should I take for this rash? ← agent must refuse with a clear reason (booking-only rule)

Add GEMINI_API_KEY in .env.local and Vercel project env, then redeploy / restart to enable the agent chat.

Agent tools

Tool name	What it does	Reads or writes
searchDoctors	Match department / language / issue to doctors	Read
getAvailableSlots	List free slots for a doctor and date	Read
bookAppointment	Book a free slot for P001	Write
rescheduleAppointment	Move a booking to another free slot	Write
cancelAppointment	Cancel a booking and free the slot	Write
getMyAppointments	List the logged-in patient’s appointments	Read

Database tables

• departments — specialties and common issues the agent can match
• doctors — clinicians, fees, languages, rooms
• slots — bookable times; is_booked enforces single booking
• patients — demo patients including P001 Arjun Mehta
• appointments — booked / cancelled / completed visits
• accounts — password hashes for patient/doctor logins (seed/accounts_auth.sql)

Tech stack

Next.js · React · TypeScript · Tailwind CSS · Supabase · Gemini API (function calling) · Zod · Vercel · JWT role-based sessions (jose)

Team

Role	Name
Team Lead	Rishi Krishna Sahoo
Frontend	Anik Das
Backend / API	Anik Das
Database	Panthadeep Majumder
AI / Agent	Jay Mukati

Run it locally

1. npm install
2. Copy .env.example to .env.local and fill in the keys (GEMINI_API_KEY, Supabase URL + secret key, APP_JWT_SECRET)
3. In Supabase SQL Editor, run seed/ps3_clinic_appointments.sql, then seed/accounts_auth.sql
4. npm run dev
5. Open http://localhost:3000 — sign in as P001 / DR01 with password demo1234, or register a new account

Auth model

1. Register stores a scrypt password hash in accounts; login verifies id + password.
2. On success the server signs an HS256 JWT (APP_JWT_SECRET) with claims: role (patient | doctor), userId, name.
3. The JWT is stored in an httpOnly cookie (careslot_session).
4. proxy.ts verifies the JWT on every page/API. Write tools always use the patient id from the JWT, never from the model.