Accessify transforms inaccessible STEM documents into screen-reader friendly content while providing AI-powered research capabilities - making complex mathematical content accessible to everyone.

Accessify is an AI-powered research assistant that combines NotebookLM-style document interaction with advanced accessibility features. It automatically converts PDF documents into accessible HTML format using Mathpix and Gemini AI, while providing chat-based document interaction, citation tracking, and comprehensive accessibility compliance.

• Document Chat: Interactive conversation with uploaded documents
• Verifiable Citations: Direct source links with line-by-line references
• Audio Generation: AI-powered podcast summaries from document content
• Multi-format Support: PDF, text, websites, YouTube videos, and audio files

• Automatic PDF Conversion: Mathpix integration converts PDFs to LaTeX-formatted HTML
• AI-Generated Alt Text: Gemini AI creates WCAG-compliant descriptions for images and figures
• Screen Reader Optimization: Mathematical equations rendered as accessible HTML
• Accessibility Scoring: Real-time accessibility compliance metrics
• Semantic HTML: Proper ARIA labels, headings, and document structure

• Real-time Processing: Live document processing with progress indicators
• Vector Search: RAG-powered document retrieval and context generation
• Multi-source Notebooks: Combine multiple document types in single research sessions
• Download & Export: Save accessible HTML versions of processed documents

• React 18 with TypeScript
• Vite for build tooling
• Tailwind CSS for styling
• shadcn/ui component library
• React Query for state management
• Supabase Client for real-time data

• Supabase - Database, authentication, and file storage
• Supabase Edge Functions - Serverless compute (Deno runtime)
• PostgreSQL - Primary database with vector extensions
• N8N Workflows - AI processing and automation

• Mathpix API - PDF to LaTeX/HTML conversion
• Google Gemini 2.5 Flash - Alt text and accessibility enhancement
• OpenAI GPT - Chat responses and content generation
• OpenAI Embeddings - Vector search and RAG implementation
• Whisper API - Audio transcription

• Supabase Storage - File upload and management
• Vector Database - Document chunk storage and similarity search
• Real-time Subscriptions - Live updates and notifications

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   Frontend      │    │   Supabase       │    │   AI Services   │
│   (React/Vite)  │◄──►│   Edge Functions │◄──►│   (Mathpix/     │
│                 │    │   (Deno)         │    │    Gemini/GPT)  │
└─────────────────┘    └──────────────────┘    └─────────────────┘
         │                       │
         │                       │
         ▼                       ▼
┌─────────────────┐    ┌──────────────────┐
│   N8N Workflows │    │   PostgreSQL +   │
│   (Automation)  │    │   Vector Store   │
└─────────────────┘    └──────────────────┘

• Node.js 18+
• Supabase account
• N8N instance (self-hosted or cloud)
• API keys for: OpenAI, Mathpix, Gemini

# Clone repository
git clone <repository-url>
cd Accessify

# Install dependencies
npm install

# Set up Supabase
npx supabase login
npx supabase link --project-ref <your-project-ref>
npx supabase db push

Create .env.local:

VITE_SUPABASE_URL=your_supabase_url
VITE_SUPABASE_ANON_KEY=your_supabase_anon_key

# Deploy all functions
npx supabase functions deploy preprocess-pdf-mathpix
npx supabase functions deploy enhance-accessibility
npx supabase functions deploy upsert-to-vector-store
npx supabase functions deploy send-chat-message
npx supabase functions deploy generate-notebook-content
npx supabase functions deploy process-document

Add to Supabase Edge Functions secrets:

GEMINI_API_KEY=your_gemini_api_key
OPENAI_API_KEY=your_openai_api_key
MATHPIX_APP_ID=your_mathpix_app_id
MATHPIX_APP_KEY=your_mathpix_app_key

Import the following workflows from /n8n directory:

• InsightsLM___Chat.json
• InsightsLM___Extract_Text.json
• InsightsLM___Generate_Notebook_Details.json
• InsightsLM___Podcast_Generation.json
• InsightsLM___Process_Additional_Sources.json
• InsightsLM___Upsert_to_Vector_Store.json

npm run dev

• POST /functions/v1/preprocess-pdf-mathpix - PDF accessibility conversion
• POST /functions/v1/enhance-accessibility - AI accessibility enhancement
• POST /functions/v1/process-document - General document processing

• POST /functions/v1/send-chat-message - Document chat interface
• POST /functions/v1/generate-notebook-content - Content generation
• POST /functions/v1/upsert-to-vector-store - Vector embeddings

• notebooks - Research session containers
• sources - Document metadata and content
• documents - Vector embeddings and chunks
• messages - Chat conversation history

• Vector Search: PostgreSQL with pgvector extension
• Real-time Updates: Supabase realtime subscriptions
• File Storage: Supabase Storage for document files
• Accessibility Metadata: WCAG compliance tracking

• Alt Text: AI-generated descriptions for all images
• Semantic HTML: Proper heading hierarchy and landmarks
• Screen Reader Support: Mathematical content as readable text
• Keyboard Navigation: Full keyboard accessibility
• Color Contrast: WCAG AA compliant color schemes

• Mathpix Conversion: PDF equations to accessible HTML
• Gemini Enhancement: AI-generated alt text and descriptions
• Semantic Structure: Proper ARIA labels and roles
• Progressive Enhancement: Works without JavaScript

1. Fork the repository
2. Create a feature branch (git checkout -b feature/accessibility-enhancement)
3. Commit changes (git commit -m 'Add accessibility feature')
4. Push to branch (git push origin feature/accessibility-enhancement)
5. Open a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

• Mathpix for PDF to LaTeX conversion capabilities
• Google Gemini for AI-powered accessibility enhancement
• Supabase for backend infrastructure
• N8N for workflow automation
• OpenAI for chat and content generation capabilities