🧠 GraphRAG Digital Twin & Interactive Portfolio

Production Monorepo: Next.js 15 (React 19, Turbopack, Tailwind CSS v4, Motion) + FastAPI Agentic GraphRAG Backend (LangGraph, Knowledge Graph, Pinecone Vector Search, LLM Guardrails).

Created by Ahmed Bargady — PhD Student in AI & Cybersecurity at UM6P (Mohammed VI Polytechnic University).

---

---

🌟 Architecture Overview

This monorepo powers an AI-driven digital twin designed to answer complex technical, research, and career inquiries in real-time. Unlike standard RAG systems, it combines Agentic Corrective RAG (LangGraph), structured Knowledge Graphs, and AI Security Guardrails.

               ┌────────────────────────────────────────────────────────┐
               │              Next.js 15 Frontend (App Router)          │
               │   Chat Interface • Knowledge Graph Viz • Security Hub   │
               └───────────────────────────┬────────────────────────────┘
                                           │ SSE / Stream
                                           ▼
               ┌────────────────────────────────────────────────────────┐
               │             FastAPI GraphRAG Server Endpoint           │
               └───────────────────────────┬────────────────────────────┘
                                           │
                        ┌──────────────────┴──────────────────┐
                        ▼                                     ▼
           ┌────────────────────────┐            ┌────────────────────────┐
           │   AI Security Shield   │            │   LangGraph State Engine│
           │ Prompt Leak & Injection│            │  Adaptive Routing Node │
           └────────────────────────┘            └───────────┬────────────┘
                                                             │
                                        ┌────────────────────┴────────────────────┐
                                        ▼                                         ▼
                           ┌────────────────────────┐                ┌────────────────────────┐
                           │   Knowledge Graph      │                │   Pinecone Vector DB   │
                           │ Multi-Hop Entity Facts │                │  Semantic Document RAG │
                           └────────────────────────┘                └────────────────────────┘

---

⚡ Core Technical Highlights

1. 🤖 Agentic GraphRAG (LangGraph)

• Adaptive Routing: Automatically detects query intent (Smalltalk, General Knowledge, Entity-Grounded, or Multi-hop Complex).
• Corrective Retrieval Loop: Grades retrieved facts; if context is insufficient, it rewrites the query dynamically and re-retrieves before generation.
• Hybrid Retrieval: Fuses deterministic Knowledge Graph subgraphs with high-dimensional vector embeddings for max recall and precision.

---

2. 🕸 Interactive Knowledge Graph

• Entity Linking: Maps research domains (APT Detection, Provenance Graphs, GNNs, Transformers) to papers, projects, and skills.
• Live Traversal Stream: Emits active graph paths alongside response streams, rendering interactive node topologies in the UI.

---

3. 🛡 AI Security Showcase & Red-Teaming Guardrails

• Prompt Injection Defense: Pre-execution input classification blocking adversarial jailbreaks.
• System Prompt Leak Guard: Output stream monitoring to prevent sensitive context disclosure.
• Rate Limiting & Red-Teaming Suite: Built-in endpoints for automated security evaluation.
• Data Privacy & Open-Source Audit: All biography, research, and project knowledge base files stored under backend/data/ are 100% sanitized, public-ready, and clean of private credentials, SSH keys, or confidential secrets.

---

📁 Repository Layout

GraphRAGPortfolio/
├── frontend/             # Next.js 15 (React 19, Turbopack, Tailwind CSS v4, Motion, Lucide)
│   ├── src/
│   │   ├── app/          # App router pages (Home, Chat, Security Showcase)
│   │   ├── components/   # Interactive Graph, Chat UI, Reasoning Trace, Terminal components
│   │   └── store/        # Zustand state management
│   └── package.json
│
├── backend/              # FastAPI Python Backend
│   ├── agent/            # LangGraph state machine & Corrective RAG pipeline
│   ├── kg/               # Knowledge graph schema, networkx, & store
│   ├── security/         # Red-teaming scripts, guardrails, & rate limiters
│   ├── data/             # Research documents, papers, & bio knowledge base
│   ├── config.py         # Centralized model & server configuration
│   └── requirements.txt
│
├── package.json          # Root npm workspace & script orchestration
├── turbo.json            # Turborepo task pipeline configuration
├── .gitignore
└── .env.example

---

🚀 Quick Start

Prerequisites

• Node.js: v20+
• npm or pnpm
• Python: v3.10+

---

1. Installation

Clone the repository and install root workspace dependencies:

git clone https://github.com/AhmedCoolProjects/GraphRAGPortfolio.git
cd GraphRAGPortfolio

# Install Node dependencies for all workspaces
npm install

Set up Python virtual environment for the backend:

cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cd ..

---

2. Environment Configuration

Copy .env.example to set up environment variables:

cp .env.example .env

Configure your environment variables:

• GROQ_API_KEY: Groq API Key for fast LLM inference.
• PINECONE_API_KEY: Pinecone Vector Store API Key.
• NEXT_PUBLIC_CHAT_API_URL: Backend URL (e.g. http://localhost:8000).

---

3. Local Development

Run both Frontend and Backend concurrently with Turborepo:

npm run dev

• Frontend: http://localhost:3000
• Backend API: http://localhost:8000
• Swagger Docs: http://localhost:8000/docs

---

📜 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

👨‍💻 Connect & Follow

Ahmed Bargady

PhD Student in AI & Cybersecurity @ UM6P