CogniSync

Academic content, simplified for you.

CogniSync is an AI-powered web application that transforms dense academic documents into clear, structured, and actionable content - tailored to how your brain actually works.

Upload a syllabus, paste assignment instructions, or drop in lecture notes. CogniSync extracts key points, builds a prioritized task list, rewrites everything in plain language, and adapts the experience to your cognitive profile.

Built for students with ADHD, dyslexia, anxiety, and other cognitive differences - but genuinely useful for anyone facing a wall of academic text.

---

Live App

Frontend: cogni-sync-delta.vercel.app

---

Features

Core

• Document ingestion - PDF, DOCX, PPTX, XLSX, TXT up to 100 MB, or paste text directly
• AI-powered simplification - Key point extraction, task generation, and plain-language rewrite in one pass via Anthropic Claude
• Adaptation profiles - ADHD, Dyslexia, Anxiety, and Default profiles that tune AI tone, sentence structure, and UI layout
• Complexity dial - Rewrite content at any reading level from Kindergarten to Graduate in real time
• Priority matrix - Tasks auto-classified into Eisenhower quadrants (Do Now / Schedule / Delegate / Eliminate)
• Reading modes - Focus View (one concept at a time) and Step-by-Step View (numbered instructions)
• Complexity scoring - Flesch-Kincaid before/after scores with percentage reduction shown visually

Organization and Sharing

• Session collections - Organize saved sessions into named collections with rename, move, and inline creation
• Share links - Encode simplified content into shareable URLs
• Calendar export - Detected deadlines exported as .ics files

Accessibility and UX

• Complexity heatmap - Sentence-level difficulty visualization (green to red)
• Glossary - Auto-generated jargon definitions with example sentences
• Weekly digest - Tasks due in the next 7 days, surfaced across all sessions
• Browser extension - Chrome and Firefox extension for in-page simplification

---

Project Structure

CogniSync/
├── cogni-sync/          # React + TypeScript frontend (Vite)
│   ├── src/
│   │   ├── components/  # UI components
│   │   ├── pages/       # Route pages
│   │   ├── hooks/       # Custom hooks (session store, collection store, TTS, etc.)
│   │   ├── context/     # App-wide state (AppContext)
│   │   ├── processor/   # AI pipeline (mock + live API, extractor, synthesizer)
│   │   ├── scorer/      # Flesch-Kincaid scoring
│   │   ├── ingestion/   # File parsing (PDF, DOCX, PPTX, XLSX, TXT)
│   │   ├── calendar/    # ICS export
│   │   ├── glossary/    # Jargon detection
│   │   ├── share/       # URL-based sharing
│   │   ├── db/          # Shared IndexedDB module
│   │   ├── readingModes/# Focus + Step-by-Step controllers
│   │   ├── styles/      # Design tokens + component CSS
│   │   └── types/       # TypeScript interfaces
│   └── __tests__/       # Property-based + unit tests
├── server/              # Node.js + Express backend (Anthropic Claude API)
├── extension/           # Browser extension (Chrome + Firefox)
└── docs/                # Project documentation and ideation

---

Getting Started

Prerequisites

• Node.js 18+
• An Anthropic API key (get one at console.anthropic.com)

Frontend

cd cogni-sync
npm install
npm run dev

Opens at http://localhost:5173.

The frontend works without the backend using a built-in mock mode. Set VITE_USE_MOCK=true in cogni-sync/.env.local to enable it.

Backend

cd server
npm install
npm start

Create a server/.env file with your API key:

ANTHROPIC_API_KEY=sk-ant-...

Runs at http://localhost:3001.

Tests

cd cogni-sync
npm test

Runs Vitest with property-based tests (fast-check, 100 iterations per property) covering scoring, ingestion, session management, collection store, reading modes, and more.

---

Tech Stack

Layer	Technology
Frontend	React 18, TypeScript, Vite
Routing	React Router v7
AI	Anthropic Claude (claude-haiku-4-5)
Backend	Node.js, Express
Hosting	Vercel (frontend), Render (backend)
Persistence	IndexedDB (localStorage fallback)
Testing	Vitest, fast-check, Testing Library
File parsing	pdfjs-dist, mammoth, jszip, xlsx

---

Accessibility

• WCAG AA contrast ratios throughout
• Full keyboard navigation
• Screen reader support via semantic HTML and ARIA landmarks
• prefers-reduced-motion respected
• No user-scalable=no - users can zoom freely
• Skip-to-main link, aria-live regions for status updates

---

Roadmap

See docs/ROADMAP.md for planned features and priorities.

---

License

MIT