AI Study Assistant

Features

• Material Management: Upload and organize study materials (PDF, TXT, MD, PPTX, PNG, JPG, JPEG), import from YouTube (transcripts) and web pages (article extraction)
• AI-Powered Quizzes: Generate adaptive quizzes with multiple difficulty levels, open-ended answer verification, and LaTeX math support
• AI Chat: Ask questions across all your materials with a context-aware chat assistant
• Knowledge Graph: Visualize connections between concepts, with search and interactive navigation
• Progress Analytics: Track learning performance over time
• Tags & Categories: Organize materials with AI-suggested tags and categories
• Intelligent Search: Find information quickly across all materials
• LaTeX Support: Render mathematical formulas in quiz questions and answers
• Dark Mode: Full light/dark theme support

Tech Stack

• Frontend: React 19, Next.js 16 (App Router), TypeScript, TailwindCSS
• Backend: Supabase (PostgreSQL, Auth, Storage), Next.js API Routes
• AI: Google Gemini (gemini-2.0-flash) primary, Groq (llama-3.1-8b-instant) fallback
• Visualization: react-force-graph-2d (knowledge graph), Recharts (analytics)
• File Processing: pdf-parse, markdown-it, officeparser (PPTX), Gemini Vision API (images)
• Content Import: youtube-transcript (YouTube), @mozilla/readability + linkedom (web pages)
• Language Detection: franc
• Math Rendering: KaTeX
• Testing: Jest + React Testing Library

AI Provider Setup

This application uses a dual-provider AI strategy for reliability:

• Primary: Google Gemini API (gemini-2.0-flash) — higher token limits, good for large quiz generation
• Fallback: Groq API (llama-3.1-8b-instant) — activated on rate limits or errors

AI logic is centralized in lib/services/AIService.ts — the file to edit when model IDs change.

Features using AI:

• Quiz Generation: Gemini only (no fallback due to Groq's TPM limit)
• Answer Verification (open-ended): Gemini primary → Groq fallback
• Chat: Gemini primary → Groq fallback over material context
• Knowledge Graph: Gemini primary → Groq fallback for concept extraction
• Tag/Category Suggestions: AI-generated metadata for uploaded materials
• Image Text Extraction: Gemini Vision API for PNG/JPG/JPEG files

Supported Sources:

• Documents: PDF, TXT, MD, PPTX
• Images: PNG, JPG, JPEG (text extraction via Gemini Vision API)
• YouTube: video transcripts
• Web pages: article text extraction
• Math: LaTeX formulas in quiz questions and answers ($...$ inline, $$...$$ block)

Rate Limit Handling:

• Automatic retry with exponential backoff (1s, 2s, 4s)
• Graceful fallback between providers
• Detailed logging for debugging

Getting API Keys:

• Google AI API: get a free-tier key at https://ai.google.dev/
• Groq API: get a free-tier key at https://console.groq.com/

Project Structure

ai-study-assistant/
├── app/                    # Next.js App Router
│   ├── auth/               # Authentication pages
│   ├── (dashboard)/        # Main app pages (dashboard, materials, quizzes, chat, graph)
│   └── api/                # API routes
├── components/             # React components (by feature)
├── lib/
│   ├── ai/                 # Prompts, context building, language detection
│   ├── auth/               # Session & password handling
│   ├── materials/          # File validation, YouTube import
│   ├── services/           # Core services (AIService, MaterialParser, QuizService, ...)
│   └── supabase/           # Supabase client configuration
├── supabase/               # Database schema and migrations
└── proxy.ts                # Auth gate for routed requests

Environment Variables

See .env.example for all required environment variables.

License

MIT