🌐 Digital Twin Overview

Welcome to the comprehensive documentation of the Agentic GraphRAG Digital Twin system. This project is a next-generation AI twin designed to ingest knowledge, construct a dynamic semantic topology, and expose interfaces for both humans and other autonomous agents.

🚀 The One-Line Pitch

A secure, agentic GraphRAG digital twin that retrieves knowledge via vector search, reasons over semantic relationships using knowledge graphs, streams tokens asynchronously, and exposes its functions as tools via the Model Context Protocol (MCP).

---

🏗️ System Architecture

Below is the conceptual architecture showing how data flows from ingestion to visual representation and agent consumption:

---

📖 Documentation Directory

Explore the technical implementation across each phase of the project:

1. RAG Foundations: Semantic indexing using HuggingFace embeddings and Pinecone vector search.
2. Agentic GraphRAG: Reasoning agents using LangGraph and semantic graphs via NetworkX.
3. Security Guardrails: Rule-based detection using Python re for PII leakage, prompt injections, and data security.
4. MCP Server: Integration of the digital twin with the Model Context Protocol for tool discovery and execution.
5. FastAPI & SSE Streaming: Asynchronous FastAPI endpoints using Server-Sent Events (SSE) and Groq Llama-3.3-70B.
6. Graph Visualization: Rendering a dynamic force-directed knowledge graph in real-time using the HTML Canvas API.
7. Frontend Application: Next.js 16 and React 19 UI containing animations, state management, and component systems.
8. Architecture Recap: A complete end-to-end review of data flow and system dynamics.