MedhaSpark AI: Personalized Learning with AI-Powered Tutoring
Discover MedhaSpark AI, an open-source platform for personalized learning roadmaps, spaced repetition flashcards, and AI-powered educational content.
VERSION 3.0 | POWERED BY CEREBRAS AI
MedhaSpark AI
Supercharge Your Learning Journey with an AI-Powered Personalized Tutor
Personalized Roadmaps
Spaced Repetition
Multi-AI Fallback
MIT Licensed • Open Source
THE PROBLEM
The Learning Crisis
❌
Students lack personalized learning paths
⚠️
Generic courses ignore individual pace & gaps
❌
Information overload with no structured progression
⚠️
No smart repetition → knowledge fades quickly
❌
Traditional education is one-size-fits-all
THE SOLUTION
MedhaSpark AI
Your Personal AI Tutor
MedhaSpark AI generates personalized learning roadmaps, spaced-repetition flashcards, AI-powered quizzes, and intelligent content analysis for ANY topic — adapting dynamically to YOUR unique learning style, pace, and knowledge gaps.
🗺️
Personalized Roadmaps
🃏
Smart Flashcards
🧩
Adaptive Quizzes
🔬
AI Content Analysis
CORE FEATURES
7 Powerful Features
🗺️
#3776AB
AI Learning Roadmaps
🃏
#7C3AED
Spaced Repetition Flashcards
🧩
#06B6D4
AI-Powered Quizzes
🔬
#4F46E5
AI Studio
🕸️
#8B5CF6
Conceptual Weaver
🧠
#3776AB
Cognitive Load Modulator
🔐
#475569
Secure Multi-Auth System
FEATURE 01
AI-Generated Learning Roadmaps
Input any topic → get a structured hierarchical learning path
Create from syllabus text or upload files (.txt, .pdf, .docx)
Auto-generates explanations, YouTube videos & exercises per subtopic
Web-enhanced: enriches content with live search results
Track progress: Not Started → In Progress → Completed
ROADMAP FLOW
How Roadmaps Work
📝
Topic Input
🤖
AI Structure Generation
📚
Subtopic Content
Explanations + Videos + Exercises
🌐
Web Enhancement
Optional live search enrichment
📊
Progress Tracking
Not Started → In Progress → Done
Powered by Cerebras AI (2100 tok/s)
FEATURE 02
Spaced Repetition Flashcards
Create manually or AI-generate from any text
SM-2 algorithm optimizes review intervals for maximum retention
Quality ratings (0–5) dynamically adjust ease factor & scheduling
Due date management ensures reviews happen at the right time
Linked to roadmap items and topics for learning context
EF' = EF + (0.1 - (5-q)(0.08 + (5-q)×0.02))
FEATURE 03
AI-Powered Quizzes
Two modes: Multiple-Choice (4 options) & Fill-in-the-Blank
Generated from YOUR learning content for maximum relevance
Automatic grading with detailed score tracking
Full question history stored for later review
Powered by Cerebras AI at 2100 tokens/second
🎯
🤖
✅
📚
⚡
Which of the following neural network architectures is primarily used for sequence-to-sequence tasks like language translation?
Convolutional Neural Networks (CNN)
Transformer Networks
Generative Adversarial Networks (GAN)
Multilayer Perceptrons (MLP)
FEATURE 04
AI Studio
Comprehensive Text Analysis
Paste any text → get instant summary & key concept extraction
ELI5 mode: Explain Like I'm 5 for instant simplification
Analogy generation for complex & abstract topics
Mermaid diagram generation for visual learners
Context-aware Q&A on your analyzed content
FEATURE 05
Conceptual Weaver
Connect New to Known
🕸️
Analyzes new text and maps concepts to YOUR existing knowledge
📚
Pulls context from your personal roadmaps and flashcards
🌉
Generates 'knowledge bridges' explaining new → known connections
🗺️
Builds a personalized knowledge graph over time
🔄
The more you learn, the smarter the connections become
FEATURE 06
Cognitive Load Modulator
Adaptive Learning Made Easy
Breaks complex content into bite-sized digestible chunks
Simplifies for different audience levels: Beginner → Expert
Dynamically identifies prerequisite concepts
Adds reflection prompts, summaries & visual diagrams
Suggests video topics for visual reinforcement
System Architecture
TECH ARCHITECTURE
HTML/CSS/JS + Jinja2 Templates
FastAPI (Async Python)
AI Providers: Cerebras / Groq / Gemini
External APIs: YouTube + Serper + Tavily
SQLAlchemy Async ORM → PostgreSQL (prod) / SQLite (dev)
Firebase Admin SDK + Google OAuth + JWT
AI PROVIDER SYSTEM
Multi-AI Fallback Architecture
API Request
🚀 Cerebras AI
PRIMARY
2100 tok/s
⚡ Groq
FALLBACK
🤖 Gemini
FALLBACK
✅ Response
FAILOVER
⏱️
Auto-failover on rate limits
🧠
In-memory cooldown tracking
🔌
OpenAI-compatible API format
📦
Cerebras SDK + HTTP fallback
TECH STACK
Technology Stack
⚙️ Backend
🔐 Auth & Security
🚀 Infrastructure
🤖 AI / ML
🧰 Utilities & Extras
Python 3.10+
FastAPI
SQLAlchemy 2.0 (async)
PostgreSQL
SQLite
Firebase Admin SDK
Google OAuth
JWT (python-jose)
bcrypt
Render.com
Gunicorn
Uvicorn
Health Checks
Cerebras AI
Groq
Gemini
httpx
cerebras_cloud_sdk
BeautifulSoup4
pypdf
python-docx
Mermaid.js
YouTube API
Serper API
Tavily API
DATABASE DESIGN
Data Models
👤
User
> id > email > firebase_uid > profile
🗺️
LearningRoadmap
> topic > json_structure > content
🃏
Flashcard
> ease_factor > interval_days > due_at
📝
QuizAttempt
> quiz_data > score > timestamp
📌
Topic
> title > description > user_id
💾
SavedResource
> url > type > topic_id
💡
CodingSuggestion
> code > language > context
📊
UserActivityLog
> action > timestamp > metadata
📓
UserNote
> content > roadmap_id > created_at
9
Core Models
Async SQLAlchemy 2.0
SECURITY
Security & Best Practices
XSS, clickjacking & MIME sniffing prevention headers
SSRF protection: URL validation + private IP blocking
CORS restricted to known origins — no wildcard + credentials
JWT with configurable expiration + bcrypt password hashing
Environment validation at startup with clear error reporting
Security headers middleware applied on every single response
DEPLOYMENT
Production Deployment
Deployed on Render.com (Python runtime)
Gunicorn + Uvicorn workers for async production serving
PostgreSQL via Aiven with SSL certificate support
Health check endpoint at /health
Environment-based config: development vs production
Auto-generated API docs at /docs (dev only)
Internet
Render.com
Gunicorn + Uvicorn
FastAPI App
Aiven PostgreSQL
API REFERENCE
RESTful API — 25+ Endpoints
🔐 AUTH ENDPOINTS
/register
/token
/auth/google
/auth/firebase/token
👤 USERS ENDPOINTS
/users/me
/users/me/focus
/users/me/stats
🗺️ ROADMAPS ENDPOINTS
/roadmaps
/roadmaps/from-syllabus
/roadmaps/from-file
/roadmaps/{id}
🃏 FLASHCARDS ENDPOINTS
/flashcards
/flashcards/review
/flashcards/due
/flashcards/count
🤖 AI ENDPOINTS
/ai/explain-topic
/ai/generate-flashcards
/ai/generate-quiz
🔬 ADVANCED AI ENDPOINTS
/ai-studio/analyze
/conceptual-weaver/analyze
/cognitive-modulator/analyze
25+ endpoints • Full OpenAPI docs • JWT authenticated
PROJECT STRUCTURE
Modular Codebase
Clean separation of concerns
medhaspark-ai/
Settings & env config
Database, auth, security
API endpoint handlers
Business logic layer
Helper functions
Jinja2 templates
CSS, JS, images
App entry point
🏗️
Layered Architecture
Clear separation between routing, logic, and data access layers.
🔌
Dependency Injection
Enhances testability and promotes loose coupling between components.
⚡
Async First
Leverages non-blocking I/O for high-performance concurrent processing.
🔒
Security by Default
Built-in protection against vulnerabilities with strict validation guidelines.
📝
Type-Safe APIs
Ensures strictly typed data exchange with comprehensive schemas.
🧪
Environment Aware
Seamless configuration management across local, dev, and production.
MedhaSpark AI
Learn Smarter, Not Harder
Built with ❤️ for learners everywhere
7
AI Features
25+
API Endpoints
9
Data Models
3
AI Providers
MIT
Licensed
github.com/yourusername/Medha-Spark-AI
⭐ Contributions Welcome!
v3.0
- ai-tutoring
- personalized-learning
- educational-technology
- open-source
- spaced-repetition
- fastapi
- artificial-intelligence