digital-me

Your AI-powered digital twin. Fork it, add your profile, deploy to Vercel — let anyone chat with your AI clone.

What is this?

A personal portfolio page with an AI chatbot that role-plays as you. It uses your profile data, experience, and training Q&A to answer questions about your background — like having a 24/7 version of yourself available for recruiters, colleagues, or anyone curious about your work.

Live demo: digital-me-test.vercel.app

Quick Start

1. Fork & Clone

git clone https://github.com/YOUR_USERNAME/digital-me.git
cd digital-me
npm install

2. Get an OpenRouter API Key

1. Go to openrouter.ai and create an account
2. Generate an API key at openrouter.ai/keys
3. (Recommended) Set a spending limit in the OpenRouter dashboard to prevent unexpected costs

cp .env.example .env.local
# Edit .env.local and add your OPENROUTER_API_KEY

3. Configure Your Profile

Edit 4 files in the config/ directory:

File	What to edit
config/profile.json	Your name, experience, skills, education, links
config/training-qa.json	Q&A pairs that teach the AI how you'd answer specific questions
config/system-prompt.md	The AI's persona instructions (usually fine as-is)
config/settings.json	AI model, rate limits, welcome message

4. Run Locally

npm run dev

Open http://localhost:3000 to see your digital twin.

5. Deploy to Vercel

Click the Deploy with Vercel button above, or:

npm install -g vercel
vercel

Set OPENROUTER_API_KEY as an environment variable in your Vercel project settings.

Configuration

Profile (config/profile.json)

{
  "name": "Your Name",
  "headline": "Your professional headline",
  "location": "City, Country",
  "experience": [
    {
      "company": "Company Name",
      "role": "Your Role",
      "duration": "2020 - Present",
      "description": "What you did",
      "achievements": ["Achievement 1", "Achievement 2"]
    }
  ],
  "skills": {
    "technical": ["TypeScript", "React"],
    "soft": ["Leadership", "Communication"]
  },
  "education": [
    {
      "institution": "University Name",
      "degree": "Degree",
      "year": "2020"
    }
  ],
  "links": {
    "github": "https://github.com/you",
    "linkedin": "https://linkedin.com/in/you"
  }
}

Training Q&A (config/training-qa.json)

These Q&A pairs teach the AI how you'd answer specific questions. The more you add, the more authentic the responses:

[
  {
    "question": "What is your greatest achievement?",
    "answer": "Your detailed answer here..."
  }
]

AI Settings (config/settings.json)

Setting	Default	Description
ai.model	google/gemini-2.0-flash-lite-001	OpenRouter model ID
ai.temperature	0.7	Response creativity (0-1)
ai.maxTokens	1024	Max response length
rateLimit.maxRequestsPerMinute	10	Per-IP rate limit
rateLimit.maxRequestsPerHour	60	Per-IP hourly limit
chat.welcomeMessage	"Hi! I'm..."	Initial chat placeholder

Browse available models at openrouter.ai/models.

Safety Features

• Jailbreak detection — blocks attempts to override the AI's role or extract system prompts
• Rate limiting — per-IP request limits to prevent abuse (in-memory, best-effort on serverless)
• OpenRouter spending limit — set a hard cost cap in your OpenRouter dashboard (recommended)

Tech Stack

• Next.js (App Router)
• Tailwind CSS + shadcn/ui
• OpenRouter (multi-model LLM gateway)
• Vercel (deployment)

Architecture

config/           -> Your profile data (JSON + Markdown)
src/app/page.tsx  -> Profile page (Server Component)
src/app/api/chat/ -> Chat endpoint (Serverless Function)
src/lib/          -> AI client, prompt builder, safety checks
src/components/   -> UI components

The chat flow:

1. User sends message to /api/chat
2. Rate limit check (IP-based)
3. Jailbreak detection (regex patterns)
4. System prompt assembled from template + profile + Q&A
5. OpenRouter API call with conversation history
6. Response returned with word-by-word typing animation

License

MIT