StudyPilot Server

A secure, full-stack multi-agent study assistant — paste your notes, get a quiz, get graded, and track your progress over time. Built for the Google × Kaggle AI Agents: Intensive Vibe Coding Capstone Project (Freestyle Track).

---

🎯 What It Does

StudyPilot AI turns raw study notes into a full learning loop:

1. Paste or upload your notes in the Study Lab
2. An agent pipeline sanitizes the input, extracts key facts, and generates a quiz
3. Take the quiz — questions are built directly from real sentences in your notes
4. Get graded instantly, with explanations that quote your original notes back to you
5. Track your progress over time on the Dashboard, with full history of every attempt

Runs completely offline in high-fidelity Simulation Mode — no API key required. Optionally connect a free Gemini API key for live LLM-powered generation.

---

🖼️ Screenshots

(Add your dashboard, quiz, and feedback screenshots here)

Dashboard	Quiz Session	Feedback Hub

---

🧠 Multi-Agent Architecture

StudyPilot AI is built on an ADK-style multi-agent system, where each agent has a single clear responsibility and hands off to the next:

 User Notes
     │
     ▼
┌─────────────────────┐
│  Notes Extractor     │  → validates, sanitizes, extracts key terms & facts
│  Agent               │
└─────────┬────────────┘
          ▼
┌─────────────────────┐
│  Quiz Generator      │  → builds fact-based questions from real note content
│  Agent               │
└─────────┬────────────┘
          ▼
┌─────────────────────┐
│  Grader / Feedback   │  → scores answers, explains mistakes using source text
│  Agent               │
└─────────┬────────────┘
          ▼
┌─────────────────────┐
│  Progress Tracker    │  → logs attempts, tracks trends, flags weak topics
│  Agent               │
└─────────────────────┘

Key Concepts Demonstrated

Concept	Implementation
Multi-Agent System (ADK)	BaseAgent class + 4 specialized sub-agents coordinating in a pipeline
MCP Server	mcp_server.py exposes agent capabilities (extract_notes, generate_quiz, grade_answers) as MCP tools over JSON-RPC
Security Features	safety_filter.py — input size validation, prompt-injection heuristics, HTML/script sanitization before any content reaches an agent
Agent Skills	Reusable skill modules (safety_filter, db_store) declared and used across agents
Deployability	Single-command local deployment via Flask, localhost:3000

---

🛠️ Tech Stack

• Backend: Python, Flask
• Frontend: HTML, CSS, vanilla JS, Chart.js
• Data Storage: Lightweight JSON file store (db.json)
• AI Integration: Optional Google Gemini API (google-genai), with full offline simulation fallback
• Protocol: Model Context Protocol (MCP) server for agent tool exposure

---

🚀 Getting Started

Prerequisites

• Python 3.9+
• pip

Installation

git clone https://github.com/Anchal-Verma04/studypilot-ai.git
cd studypilot-ai
pip install -r requirements.txt

Run the app

python src/server.py

Then open your browser at:

http://localhost:3000

That's it — no API key needed. The app runs fully offline in Simulation Mode.

(Optional) Enable live Gemini AI

1. Get a free API key at aistudio.google.com/apikey
2. Copy .env.example to .env
3. Add your key:

   GEMINI_API_KEY=your_key_here
   

4. Restart the server

---

📂 Project Structure

studypilot-ai/
├── src/
│   ├── agents/
│   │   ├── base_agent.py
│   │   ├── notes_extractor_agent.py
│   │   ├── quiz_generator_agent.py
│   │   ├── grader_agent.py
│   │   └── progress_tracker_agent.py
│   ├── skills/
│   │   ├── safety_filter.py
│   │   └── db_store.py
│   ├── public/
│   │   ├── index.html
│   │   ├── styles.css
│   │   └── app.js
│   ├── mcp_server.py
│   ├── server.py
│   └── test_agents.py
├── db.json
├── requirements.txt
├── .env.example
└── README.md

---

🔒 Security Notes

• All user-submitted notes pass through size validation and prompt-injection heuristics before reaching any agent
• Content is HTML-sanitized to prevent script injection in the UI
• No credentials are hardcoded — API keys are loaded from a local .env file (never committed to version control)

---

🏆 About This Project

Built as the capstone project for Google & Kaggle's 5-Day AI Agents: Intensive Vibe Coding Course (Freestyle Track), demonstrating multi-agent orchestration, MCP server design, agent skills, and security-conscious agent architecture — developed using Antigravity IDE.

---

📄 License

MIT License — free to use, modify, and learn from.