Smart AI Study Assistant — Multi-Agent LLM System

A Generative AI Capstone Project for the Agents for Good Track
LLM Agents • Parallel Agents • Memory • Logging • Long-running Tasks

---

🎓 A complete offline multi-agent learning companion powered by simulated LLM intelligence.

---

Smart AI Study Assistant — Multi-Agent LLM System

A Generative AI Capstone Project (Agents for Good — Education)

This project implements a fully offline, multi-agent study assistant that simulates an intelligent AI tutor.
It demonstrates the key agent features required for the GenAI Capstone:

• LLM-powered agent (mocked for offline use)
• Parallel agents (content search + flashcard generator)
• Sequential agents (quiz generation, guidance)
• Session & memory (InMemorySessionService)
• Observability (logging)
• Long-running agent simulation (pause & resume)

---

🚀 Features

🔹 1. LLM Agent (Mocked)

Generates:

• explanations
• quizzes
• summaries
• prerequisites, key concepts, practice problems

🔹 2. Parallel Agents

Using ThreadPoolExecutor:

• ContentSearchAgent
• FlashcardAgent

🔹 3. Sequential Agents

The StudyCoordinator orchestrates:

• parallel work
• then sequential quiz generation
• then informative Q/A

🔹 4. Memory System

InMemorySessionService stores:

• user metadata
• study plan
• topics
• weekly hours

🔹 5. Logging / Observability

All agents log:

• start/finish
• progress
• warnings
• pause/resume

🔹 6. Long-running Operation Agent

Simulates:

• step-wise computation
• pausing
• resuming

---

🏛 Architecture Diagram

User
↓
StudyCoordinator
├── LLM Agent (Mock/Gemini-ready)
├── ContentSearchAgent
├── FlashcardAgent
├── LongRunningAgent
└── Session Service

---

📁 Repository Structure

├── README.md
├── LICENSE
├── requirements.txt
├── notebook/
│ └── smart_ai_study_assistant.ipynb
|  
src/
│
├── models/
│   ├── learning_objective.py
│   ├── resource.py
│   └── flashcard.py
│
├── memory/
│   └── session_service.py
│
├── agents/
│   ├── llm_agent.py
│   ├── search_agent.py
│   ├── flashcard_agent.py
│   └── coordinator.py
│
├── long_running_agent.py
│
└── main.py

---

📄 Notebook

The full implementation and demo run are inside:

This notebook includes:

• step-by-step logic
• agent orchestration
• parallel execution demo
• session memory handling
• observability via logging
• long-running agent pause/resume demonstration

---

🛠 Technologies Used

• Python 3
• Dataclasses
• concurrent.futures
• Logging (Observability)
• Mock LLM Tool
• In-memory session service

No external API calls — fully offline.

---

🧪 Demo Output

The demo generates a full study plan across 5 topics:

• curated resources
• LLM-powered flashcards
• quizzes
• prerequisites
• key concepts
• practice problems
• session memory
• agent logs
• long-running agent output

This demonstrates a working multi-agent system in action.

---

🔮 Future Enhancements

• Integrate real LLMs (Gemini/GPT)
• Add vector memory (FAISS/Chroma)
• Build a recommendation engine for next best topics
• Implement spaced repetition algorithms
• Add FastAPI backend or a Streamlit UI
• Build a prerequisite graph to generate learning paths

---

📜 License

MIT License

---

👤 Author

Kuldeep Padshala
Your GitHub Profile