MindEase — AI-Assisted Mental Wellbeing Chatbot

A simple, professional psychology website with a Gemini-powered chatbot that provides supportive, educational responses for mental wellbeing.

Features

• 🤖 AI-Powered Chat: Gemini 1.5 Flash integration for empathetic, educational guidance
• 🛡️ Safety Guardrails: Crisis keyword detection with emergency resource guidance
• 🎨 Calm UI: Soft blue/white palette with accessible design
• ⚡ Rate Limiting: Basic IP-based request throttling
• 🚀 Production Ready: Optimized for Vercel deployment

Tech Stack

• Framework: Next.js 15 (App Router)
• Language: TypeScript
• Styling: Tailwind CSS
• LLM SDK: @google/generative-ai
• Runtime: Node 18+

Getting Started

Prerequisites

• Node.js 18 or higher
• A Google Gemini API key (Get one here)

Installation

1. Clone the repository

   git clone <your-repo-url>
   cd mindease

2. Install dependencies

   npm install

3. Configure environment variables

   Create a .env.local file in the root directory:

   GEMINI_API_KEY=your_api_key_here
   GEMINI_MODEL=gemini-1.5-flash
   SITE_NAME=MindEase

4. Run the development server

   npm run dev

5. Open your browser

   Visit http://localhost:3000

Project Structure

/mindease
  /app
    /api
      /chat
        route.ts          # POST /api/chat endpoint
    globals.css
    layout.tsx
    page.tsx             # Main landing page
  /components
    Chatbot.tsx          # Chat widget UI
    CrisisNotice.tsx     # Emergency banner
  /lib
    gemini.ts            # Gemini client + system prompts
    safety.ts            # Crisis detection
    rateLimit.ts         # Simple IP rate limiter
  /public
  .env.local             # Environment variables (create this)
  package.json
  tailwind.config.ts
  tsconfig.json

API Contract

POST /api/chat

Request:

{
  "message": "I feel overwhelmed at work.",
  "sessionId": "optional-string",
  "locale": "en"
}

Response (Normal):

{
  "reply": "It sounds like you're carrying a lot...",
  "safety": {
    "isCrisis": false,
    "matched": null
  },
  "meta": {
    "model": "gemini-1.5-flash",
    "latencyMs": 742
  }
}

Response (Crisis Detected):

{
  "reply": "I'm really sorry you're going through this...",
  "safety": {
    "isCrisis": true,
    "matched": "suicide"
  },
  "meta": {
    "model": null,
    "latencyMs": 0
  }
}

Safety Features

Crisis Detection

The chatbot includes server-side crisis keyword detection that triggers before calling the AI model. Keywords include:

• suicide
• kill myself
• self-harm / self harm
• end my life
• i want to die
• don't want to live
• overdose
• cutting

When detected, the bot responds with:

• Empathetic acknowledgment
• Emergency service guidance
• Encouragement to reach out to professionals
• Clear disclaimer about educational purpose

Non-Clinical Guardrails

The system prompt ensures:

• No diagnosis or treatment plans
• No medication advice
• Empathetic, non-judgmental tone
• Short, actionable self-care suggestions
• Reminders to seek professional help

Deployment

Vercel (Recommended)

1. Push your code to GitHub
2. Import the repository in Vercel
3. Add environment variables in Vercel dashboard:
   • GEMINI_API_KEY
   • GEMINI_MODEL (optional, defaults to gemini-1.5-flash)
4. Deploy

Other Platforms

This app can run on any Node.js hosting platform. Ensure:

• Node 18+ runtime
• Environment variables are configured
• Build command: npm run build
• Start command: npm start

Testing

Manual Testing Checklist

• Normal message returns empathetic AI response
• Crisis phrase triggers emergency response (no AI call)
• Empty message returns 400 error
• Enter key submits message
• Send button disabled when input is empty
• Messages auto-scroll to bottom
• Rate limiting works after multiple requests
• Error messages display clearly

Test Crisis Detection

Send a message containing crisis keywords:

"I want to kill myself"
"I'm thinking about self-harm"

Expected: Emergency response, no AI model call, crisis notice displayed.

Test Normal Flow

Send a regular message:

"I feel stressed about work"

Expected: Empathetic, educational AI response.

Important Notes

⚠️ This is an educational tool, not a medical service.

• Does not replace professional mental health care
• Not suitable for crisis intervention
• Always directs users in crisis to emergency services
• No user data is stored server-side

Future Enhancements

• Multi-language support (ES/EN)
• FAQ and resources pages
• Dark mode theme toggle
• Client-side chat history (localStorage)
• Advanced analytics and logging
• User accounts (optional)

License

ISC

Support

For questions or issues, please open a GitHub issue or contact the maintainers.

---

Remember: If you or someone you know is in crisis, please contact local emergency services immediately.