IntelliStudy — Advanced AI-Powered Learning Ecosystem 🧠🚀

IntelliStudy is a state-of-the-art Contextual Enrichment Framework (CEF) designed for academic and professional document interaction. Built with Java 21 and Spring Boot 3.5, it integrates Graph-Augmented RAG (Retrieval-Augmented Generation) to provide deep semantic insights, automated assessments, and linguistic bridging.

---

🛠️ Performance & Technology Stack (Why & What?)

The technical choices behind IntelliStudy are mission-driven, prioritizing high-precision retrieval and secure scalability.

Technology	Implementation	Why this was chosen?
Java 21	Backend Core	Modern syntax, Pattern Matching, and Virtual Threads readiness for high throughput.
Spring Boot 3.5	Framework	Industry-standard for robust, maintainable, and highly decoupled micro-services.
Gemini 2.5 Flash	Core LLM Engine	Superior speed-to-intelligence ratio and massive context window for long docs.
DuckDB	Vector DB	In-process analytical database allowing blazingly fast local vector search (768d).
Redis	Metadata Cache	Eliminates repetitive AI calls by caching RAG contexts and session history.
JGraphT	Semantic Graph	Enables BFS-based context expansion, solving the "lost-in-the-middle" problem.
JWT + Rotation	Security	Ensures statelessness while maintaining high security via DB-backed token rotation.
Docker	Orchestration	Guaranteeing "Works on My Machine" (WOMM) parity across environments.

---

🏗️ The Engineering Architecture

At its core, IntelliStudy implements a Clean Layered Architecture powered by our proprietary CEF engine:

1. Mcp (Model Context Protocol) Routing Layer

The McpAgentLayer acts as the "Brain" of the system. Instead of simple prompting, it uses Intent Classification to route queries to specialized agents:

• Intelligent Dispatcher: Uses gemini-2.5-flash to classify user intent into discrete Tools (RAG_ASK, SUMMARIZE, QUIZ, TRANSLATE).
• Session-Awareness: Maintains context within the CEF layer for multi-turn document sessions.

2. The RAG Pipeline (Graph + Vector)

IntelliStudy uses a hybrid retrieval strategy combining semantic vector search with graph-based semantic traversal:

• Vector Search (DuckDB): Text is split into chunks (1000 chars, 200 overlap) and embedded via text-embedding-004.
• GraphRAG (JGraphT): Maps semantic relationships (Nodes/Edges) between extracted entities.
• Graph-Boosted Scoring: During retrieval, if a chunk contains a neighbor node (BFS Depth=2) of the query concept, it receives a +0.15 score boost.

$$FinalScore = VectorRankScore + \sum(GraphMatchBoost)$$

3. Identity & Persistence Layer

• JWT + Refresh Token Rotation: Secure stateless sessions with hashed MySQL-backed refresh identifiers.
• Google OAuth2: Seamless integration for Social Identity Providers.

---

✨ Features Walkthrough

• 🧠 Intelligent Q&A: Graph-boosted RAG ensures answers are contextually accurate and topologically linked.
• 📄 Pro Summarization: Multi-stage extraction designed for high-density academic papers (12k char window).
• 🇧🇩 English-Bangla Bridge: High-fidelity translation maintaining domain-specific terminology.
• 📝 MCQ Generator:
  • Dynamic Topic Mode: Generates quizzes from a single keyword using semantic knowledge.
  • Contextual Text Mode: Extracts specific "factoids" from your uploaded PDF using AI.
• 🎓 Interactive Assessment: Built-in score evaluation with semantic feedback and correct answer reviews.

---

🚀 Deployment Guide (Dockerized)

1. Clone the repository:

   git clone https://github.com/naim79992/IntelliStudy.git
   cd IntelliStudy

2. Configure your environment:
   • Create a .env file and add your GEMINI_API_KEY.
3. Run the stack:

   docker-compose up --build -d

4. Access the dashboard: http://localhost:8080

---

🖼️ Gallery & UI Documentation

🏠 System Home

🧬 AI-Driven Quiz (Topic Specific)

📊 Assessment Architecture (Text Based)

📑 Contextual Intelligence (RAG)

---

Architected for Intelligence. Built for Learning.
Developed by Naim .