Docuer

Inspiration

Information overload is real. We noticed that learning from dense technical documentation feels overwhelming and boring for most people.
TikTok changed how we consume content. Why not apply the same bite-sized, swipeable approach to learning?
Everyone learns differently. We wanted to create a system that adapts to your skill level, interests, and goals instead of forcing a one-size-fits-all approach.
Documentation shouldn't gather dust. We saw an opportunity to transform static docs into interactive, personalized learning experiences.

Transforms any documentation website into bite-sized courses - Just paste a URL and watch it become a structured learning path
Generates personalized 160-word articles tailored to your experience level (beginner, intermediate, advanced)
Creates interactive knowledge graphs showing how topics connect and what prerequisites you need
Adapts learning paths in real-time based on your profile, interests, and goals
Tracks your progress intelligently - completion rates, quiz scores, learning patterns, and even what topics you struggle with
Imports from Google Drive so you can learn from your own documents
Tests your knowledge with auto-generated quizzes that match your difficulty level
Provides an AI chat assistant to answer questions while you learn
TikTok-style interface - swipe through lessons on mobile, making learning feel effortless

Frontend: Next.js 16 with React 19, HeroUI components, Tailwind CSS for that clean look, and Framer Motion for smooth animations
AI Brain: Google Gemini 2.0 Flash for content generation, topic extraction, and personalization (with built-in rate limiting so we don't blow through API calls)
Knowledge Graph: Neo4j database to store course structures, relationships, and learning paths - perfect for understanding how concepts connect
Memory System: Supermemory for behavior tracking and documentation caching with hash-based deduplication (multiple users share cached docs = cost efficient)
Web Scraping: Firecrawl to intelligently crawl documentation sites and extract clean content
Smart Architecture: Two-phase content selection (AI recommends pages based on your profile, then you choose), semantic relationship detection, and dangling node auto-connection for fully connected graphs
State Management: Zustand with localStorage persistence so your progress never gets lost

Rate limiting nightmares - Gemini has strict limits, so we built throttling logic to queue requests and prevent API errors
Knowledge graph complexity - Making sure topics connect meaningfully without creating isolated nodes took multiple iterations
Personalization accuracy - Balancing difficulty matching (35%), interests (25%), goals (25%), and importance (15%) to generate truly relevant learning paths
Topological sorting for prerequisites - Ensuring learners don't hit advanced topics before covering the basics
Hash-based deduplication - Figuring out how to share cached documentation across users without privacy issues
Mobile-first design - Making a TikTok-style interface work smoothly while keeping desktop usable
Google Drive OAuth flow - Dealing with COOP (Cross-Origin-Opener-Policy) restrictions and callback handling
Content quality control - Ensuring AI-generated articles are accurate, concise, and actually helpful

Nailed the personalization - The AI genuinely adapts content to different skill levels and learning styles
Hash-based sharing works beautifully - Multiple students studying React docs share the same cached content, saving time and money
Knowledge graphs look amazing - The force-directed visualization makes complex topics feel manageable
Two-phase crawling is genius - AI recommends pages before crawling, so users don't waste time on irrelevant content
Behavioral analytics that actually help - The system identifies struggling areas and preferred topics automatically
Built-in rate limiting - No more API errors or unexpected bills
Clean, intuitive UI - Users immediately understand how to navigate and learn
Zero-config fallbacks - Works with just localStorage if you don't have Neo4j or Supermemory set up

AI personalization is powerful but tricky - Getting the scoring weights right took experimentation
Graph databases are perfect for learning paths - Neo4j made prerequisite tracking and semantic relationships so much easier than traditional SQL
Rate limiting is non-negotiable - Always build throttling into AI integrations from day one
Bite-sized content really works - 160 words feels perfect for mobile learning
Caching is critical - Hash-based deduplication saved us tons of API calls
User behavior data is gold - Tracking interactions reveals learning patterns we never expected
Fallback strategies matter - Always have mock data and error handling ready
Modern React is fast - Next.js 16 + React 19 gave us incredible performance

Production-ready authentication - Replace prototype auth with NextAuth.js or OAuth
Spaced repetition system - Automatically resurface topics before you forget them
Real-time collaboration - Study with friends, share notes, compete on leaderboards
Video content integration - Link YouTube tutorials to relevant articles
Mobile app - Native iOS/Android experience with offline mode
Advanced analytics dashboard - Visualize learning velocity, knowledge retention, and skill gaps
Plugin system - Let developers add custom content sources (Notion, Confluence, GitHub wikis)
Social features - Share courses, follow other learners, showcase completed paths
Multi-language support - Translate content automatically for global learners
Enterprise features - Team management, custom branding, progress reporting for organizations
API marketplace - Let third parties build on top of Docuer's learning engine
Gamification - Badges, streaks, achievements to keep learners motivated

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.