Personal Learning Intelligence - An adaptive AI tutor powered by Google Gemini 3

EducraAI leverages the full power of Gemini 3:

Educra AI is a next-generation learning companion that goes beyond simple Q&A chatbots. It combines:

• Real-time Voice Tutoring with Gemini Live API
• Adaptive Confusion Detection that senses when you're struggling
• Living Notes that auto-generate as you learn
• Knowledge Graph Visualization to see your understanding
• Cognitive-Science-Based Quizzing with spaced repetition
• Proactive Learning Interventions based on memory decay

flowchart TB
    subgraph Frontend["Frontend (Next.js 15)"]
        UI[TutorSession - Voice/Text Chat]
        KG[Knowledge Graph Viz]
        LN[Living Notes]
        QM[Quiz Mode]
        MS[Mind Stream]
        PB[Proactive Bar]
    end
    
    subgraph Backend["Backend (FastAPI + Python)"]
        AS[Agent Server]
        GA[Gemini Agent<br/>Thinking Models]
        GLA[Gemini Live Agent<br/>Voice Streaming]
        EA[Evaluator Agent]
        EMA[Episodic Memory Agent]
        PE[Proactive Engine]
        KGA[Knowledge Graph Agent]
        CA[Consolidation Agent]
    end
    
    subgraph External["Gemini 3 APIs"]
        G3[Gemini 3 Flash<br/>+ Google Search Grounding]
        GL[Gemini Live API<br/>Real-time Audio]
    end
    
    subgraph Storage["Persistence"]
        N4J[(Neo4j Aura<br/>Knowledge Graph)]
        LS[(LocalStorage<br/>Client State)]
    end
    
    UI -->|REST/WebSocket| AS
    KG -->|Concepts| AS
    LN -->|Episodes| AS
    MS -->|Recall| AS
    PB -->|Interventions| AS
    QM -->|Evaluation| AS
    
    AS --> GA
    AS -->|WebSocket| GLA
    AS --> EA
    AS --> EMA
    AS --> PE
    AS --> KGA
    AS --> CA
    
    GA -->|Thinking + Grounding| G3
    GLA -->|Bidirectional Audio| GL
    EMA --> N4J
    KGA --> N4J
    
    UI --> LS
    KG --> LS

Real-time bidirectional audio with Gemini Live. Speak naturally and get instant voice responses.

AI automatically extracts key concepts and generates study notes as you learn. No more manual note-taking.

The AI senses when you're confused based on your questions and adapts its teaching approach in real-time.

Visualize your learning progress across concepts. See what you've mastered and what needs work.

Based on cognitive science research:

• Spaced Repetition - Review at optimal intervals
• Retrieval Practice - Active recall strengthens memory
• Desirable Difficulty - Challenging questions promote deeper learning
• Adaptive Difficulty - Questions adjust to your level

Generate beautiful diagrams (mind maps, flowcharts, hierarchies) for any concept using AI.

Import your existing study notes from Notion to enhance personalized learning.

Get gentle nudges for spaced repetition reviews based on memory decay algorithms.

View your complete learning journey with episodic memory. Every session is stored with linked concepts, timestamps, and source types (voice, text, notes).

Research-backed progress tracking with:

• Mastery Progress - Track concept understanding
• Active Recall Score - Questions asked and quiz performance
• Feynman Readiness - Concepts ready to teach
• Review Streak - Spaced repetition adherence
• Expandable Research Citations - Learn about the science behind your learning

• Visual Audio Waveform - See your voice as you speak
• AI Speaking Indicator - Know when the AI is responding
• Recording Status - Clear feedback during voice sessions

• Node.js 18+
• Python 3.11+
• Gemini API Key (Get one here)

cd EducraAI

# Frontend
npm install

# Backend
cd backend
pip install -r requirements.txt

Create .env.local in the root:

NEXT_PUBLIC_GEMINI_API_KEY=your_gemini_api_key
GEMINI_API_KEY=your_gemini_api_key

# Terminal 1: Backend
cd backend
python agent_server.py

# Terminal 2: Frontend
npm run dev

Open http://localhost:3000

EducraAI/
├── src/
│   ├── app/                 # Next.js App Router
│   │   └── page.tsx         # Main dashboard
│   ├── components/
│   │   ├── TutorSession.tsx      # Voice/text tutoring
│   │   ├── LivingNotes.tsx       # Auto-generated notes
│   │   ├── KnowledgeGraphViz.tsx # ReactFlow visualization
│   │   ├── QuizMode.tsx          # Cognitive quiz engine
│   │   ├── ConceptVisualizer.tsx # Mermaid diagrams
│   │   ├── NotionIntegration.tsx # Notion import
│   │   ├── MindStream.tsx        # Episodic memory feed
│   │   ├── ProactiveBar.tsx      # Learning reminders
│   │   └── ConfusionMeter.tsx    # Understanding gauge
│   └── lib/
│       ├── gemini-live.ts        # Gemini Live client
│       ├── gemini-socket.ts      # WebSocket handler
│       └── knowledge-graph.ts    # KG state management
├── backend/
│   ├── agent_server.py           # FastAPI main server
│   └── agents/
│       ├── gemini_agent.py       # Text chat agent
│       ├── gemini_live_agent.py  # Voice agent
│       ├── evaluator_agent.py    # Answer verification
│       ├── episodic_memory_agent.py  # Learning memory
│       ├── proactive_engine.py   # Intervention logic
│       └── knowledge_graph_agent.py  # Concept tracking
└── public/

Educra AI is built on established cognitive science research:

• Next.js 15 - React framework with App Router
• TailwindCSS - Utility-first styling
• Framer Motion - Animations
• ReactFlow - Knowledge graph visualization
• Mermaid - Diagram generation

• FastAPI - High-performance Python API
• Google GenAI SDK - Gemini 3 integration
• Neo4j (optional) - Graph database for knowledge persistence

• Gemini 3 Flash - Thinking model with Google Search grounding
• Gemini Live API - Real-time bidirectional audio

Contributions are welcome! Please read our contributing guidelines first.

MIT License - see LICENSE for details.

Built with ❤️ for the Google Gemini 3 Hackathon 2026

Educra AI - Learn smarter, not harder.