TheaMarieM/digital-twin-portfolio

Digital Twin Portfolio

AI-Powered Portfolio with MCP Integration, RAG System, and Production-Ready Security

A Next.js portfolio application featuring a digital twin AI assistant powered by Claude Desktop, Groq LLaMA 3.3, and semantic search capabilities. Built with enterprise-grade security, rate limiting, and comprehensive testing.

Features

Modern UI/UX

• Gradient Theme: Purple-to-pink gradient design with glassmorphism effects
• Dark Mode: Full theme toggle support with CSS custom properties
• Responsive Design: Mobile-first approach with adaptive layouts
• Custom Animations: Smooth transitions, hover effects, and micro-interactions
• Floating Cat Assistant: Animated cat button for AI chat navigation
• Professional Footer: Social links with gradient hover effects

AI & MCP Integration

• Claude Desktop Integration: OPERATIONAL - Full MCP protocol support
• 7 Production Tools: Portfolio queries, skills assessment, interview prep, RAG search
• Digital Twin Assistant: Context-aware responses about professional experience
• FastAPI MCP Server: HTTP and stdio communication modes

RAG System

• Semantic Search: Upstash Vector with 16 embedded portfolio chunks
• OpenAI Embeddings: text-embedding-3-small for high-quality vectors
• Context Retrieval: Intelligent portfolio data extraction
• Local Embeddings: Optional sentence-transformers (384-dim) for development

Enterprise Security

• Rate Limiting: Redis-based distributed rate limiting (20 req/min chat, 30 req/min RAG)
• Input Validation: Comprehensive message/query sanitization
• Content Filtering: API key detection, password filtering, prompt injection prevention
• Crypto-Secure Sessions: 256-bit secure session ID generation
• Security Headers: X-Content-Type-Options, X-Frame-Options, X-XSS-Protection
• Session Management: 100 messages per session limit with automatic cleanup

Data Management

• Upstash Redis: Chat history, session storage, rate limiting
• STAR Format: 5 structured professional experience items
• Interview Q&A: 10 pre-generated interview training pairs
• Semantic Cache: Cosine similarity (0.95 threshold), 100-item capacity, 30-min TTL

Quick Start

Prerequisites

• Node.js 18+ and pnpm installed
• Python 3.8+ with pip
• Upstash accounts (Redis + Vector)
• Groq API key (or OpenAI for embeddings)

Installation

# Clone the repository
git clone https://github.com/TheaMarieM/digital-twin-portfolio.git
cd digital-twin-portfolio

# Install Node.js dependencies
pnpm install

# Create Python virtual environment
python -m venv .venv
.venv\Scripts\Activate.ps1  # Windows PowerShell

# Install Python dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env.local
# Edit .env.local with your API keys

Environment Variables

Create a .env.local file with the following:

# OpenAI (for embeddings)
OPENAI_API_KEY=sk-...

# Groq (for LLM responses)
GROQ_API_KEY=gsk_...

# Upstash Vector
UPSTASH_VECTOR_REST_URL=https://...upstash.io
UPSTASH_VECTOR_REST_TOKEN=...

# Upstash Redis
UPSTASH_REDIS_REST_URL=https://...upstash.io
UPSTASH_REDIS_REST_TOKEN=...

Running the Application

# Start Next.js dev server
pnpm dev
# Open http://localhost:3001

# Start MCP server (optional, for Claude Desktop)
python mcp/server.py
# Server runs on http://127.0.0.1:3000

Architecture

my-ai-app/
 app/                        # Next.js App Router
    api/
       chat/              # AI chat endpoint (Groq + LLaMA 3.3)
       rag/               # RAG semantic search endpoint
       metrics/           # Performance metrics
    components/
       sections/          # Page sections (Hero, Skills, Contact)
       ui/                # Reusable UI components
    page.tsx               # Home page with all sections
 mcp/                        # Model Context Protocol Server
    server.py              # FastAPI HTTP server (port 3000)
    stdio_server.py        # stdio MCP server for Claude Desktop
    tools/                 # MCP tool implementations
        portfolio_tools.py # Portfolio queries, projects, skills
        interview_tools.py # Interview Q&A, question generation
        rag_tools.py       # Semantic search, embeddings
 data/
    profile.json           # 5 STAR-format portfolio items
    interview_qa.jsonl     # 10 interview training Q&A pairs
 scripts/
    chat_with_ollama.py    # CLI chat with RAG
    index_local_embeddings.py # Vector database upserter
 docs/
     CLAUDE_DESKTOP_SETUP.md # MCP setup for Claude Desktop
     VERCEL_DEPLOY.md        # Vercel deployment guide

MCP Server

Available Tools

Security Features

Rate Limiting

• Chat API: 20 requests/minute (Redis-based, distributed)
• RAG API: 30 requests/minute (in-memory, per-session)

Input Validation

• Message Length: 500 characters max for chat, 300 for RAG
• Session Messages: 100 messages per session limit
• Content Filtering: API key detection, password filtering, SQL injection prevention

Deployment

Vercel

# Install Vercel CLI
npm i -g vercel

# Deploy
vercel --prod

# Set environment variables:
vercel env add OPENAI_API_KEY
vercel env add GROQ_API_KEY
vercel env add UPSTASH_VECTOR_REST_URL
vercel env add UPSTASH_VECTOR_REST_TOKEN
vercel env add UPSTASH_REDIS_REST_URL
vercel env add UPSTASH_REDIS_REST_TOKEN

Documentation

• : MCP integration guide
• : Deployment instructions
• : Copilot coding instructions

Contact

Thea Marie

• GitHub: @TheaMarieM
• LinkedIn: linkedin.com/in/theamarie