DocuMentor AI

Learn smarter, not harder. DocuMentor AI is your adaptive reading assistant that personalizes every doc, insight, and explanation to you.

Comment

Inspiration

Developers spend countless hours in documentation, but every doc is written for a generic audience. We wanted to transform static documentation into a personalized learning experience that adapts to each developer's skill level, learning style, and goals—without breaking their flow.

What it does

DocuMentor AI is a Chrome extension that acts as your personal learning companion for technical documentation. It uses Chrome's built-in AI (Gemini Nano) to:

  • Analyze pages and determine if they match your skill level
  • Generate instant cheat sheets for quick reference
  • Explain complex concepts (text and images) in terms you understand
  • Recommend learning resources personalized to your goals
  • Track your progress and adapt as you grow

How we built it

  • Frontend: React + TypeScript for the side panel UI
  • Chrome APIs: Leveraged all three built-in AI APIs:
    • Summarizer API for content summarization
    • Writer API for structured cheat sheet generation
    • Prompt API for personalized analysis, recommendations, and multimodal image understanding
  • Architecture: Content scripts detect user interactions (text selection, image hover), service worker manages lifecycle, and the side panel orchestrates AI processing
  • Persona System: Custom user profile serialized as sharedContext for consistent AI personalization across all features

Challenges we ran into

  • Combining multiple AI APIs: Figuring out when to use Summarizer vs. Prompt vs. Writer, and orchestrating them for seamless UX
  • Multimodal limitations: Working with Prompt API's image support and converting base64 to Blobs for proper processing
  • Context management: Maintaining conversation history for follow-up questions while keeping responses personalized
  • Performance: Balancing local AI processing speed with user experience expectations
  • Persona design: Creating a profile system flexible enough for beginners to experts without overwhelming users

Accomplishments that we're proud of

  • First to use all three Chrome AI APIs together in a cohesive product
  • Privacy-first: 100% local AI processing with zero external API calls (except optional YouTube search)
  • Multimodal innovation: Successfully implemented image explanations for diagrams and architecture charts
  • Adaptive personalization: Every AI feature genuinely adapts to user context, not just generic responses
  • Clean UX: Made complex AI capabilities feel simple and accessible

What we learned

  • Chrome's built-in AI is production-ready and surprisingly capable for local processing
  • Combining multiple AI models (summarizer, writer, prompt) unlocks capabilities beyond using just one
  • Personalization context makes AI exponentially more useful—generic responses don't cut it
  • Multimodal understanding opens new possibilities for technical learning (diagrams, screenshots, charts)
  • Privacy-first AI can deliver excellent UX without compromising user data

What's next for DocuMentor AI

  • Hybrid AI mode: Choose between on-device (Chrome's Gemini Nano) for privacy or cloud models (GPT-5, Gemini 2.5 Pro) for more powerful analysis
  • Learning path tracking: Visualize your documentation journey and knowledge gaps
  • Spaced repetition: Surface old cheat sheets at optimal intervals for retention
  • Code playground integration: Test examples directly from docs without leaving the page
  • Team personas: Share learning profiles and cheat sheets with teammates
  • Cross-platform: Expand to other browsers as they adopt built-in AI APIs
  • Offline mode: Pre-download AI models for completely offline documentation learning

Built With

Share this project:

Updates