AI-Powered PDF Accessibility

Make PDFs accessible to screen readers — automatically

Supports PDF and LaTeX input — outputs accessible PDF or modified LaTeX source.
Uses AI vision to generate alt-text for images, charts, and math formulas.
Adds semantic structure (headings, paragraphs, reading order) for screen reader navigation.
Validates against PDF/UA standard with a 0-100 accessibility score.

Turn hours of manual remediation into minutes.

Quick Start • Features • API Reference • Deployment • Contributing

---

About

This project was developed in collaboration with MorphMind.

Try it now — no setup required: agentlab.morphmind.ai (Free trial available)

---

🚀 Quick Start

Option 1: Use via MorphMind AgentLab (Recommended)

The easiest way — no setup required:

1. Visit agentlab.morphmind.ai
2. Find the PDF Accessibility Remediator agent
3. Upload your PDF and get instant results

Free trial available — start remediating PDFs in seconds.

Option 2: Local Development

# Clone the repository
git clone https://github.com/AIScientists-Dev/AI-Powered-PDF-Accessibility.git
cd AI-Powered-PDF-Accessibility

# Create virtual environment
python3.11 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Install veraPDF (required for validation)
# macOS:
brew install verapdf
# Linux: Download from https://verapdf.org/software/

# Set up environment
echo "GEMINI_API_KEY=your_api_key_here" > .env

# Run the server
python -m uvicorn http_server:app --host 0.0.0.0 --port 8080

# Test it
curl http://localhost:8080/health

---

✨ Features

🎯 MorphMind Accessibility Score

Get an instant 0-100 score based on PDF/UA compliance, with letter grades and actionable insights.

{
  "score": 82,
  "grade": "B",
  "issues_by_severity": {
    "critical": 0,
    "serious": 3,
    "moderate": 1
  }
}

🤖 AI-Powered Alt-Text

Automatically generate descriptive alt-text for images and figures using Google Gemini vision AI.

📋 25 Specialized Tools

Category	Tools
Analysis	analyze_pdf, validate_pdfua, validate_pdfa, detect_headings
Remediation	make_accessible, add_full_structure, add_alt_text, fix_link_alt_texts
Figures	extract_figures, generate_alt_text, get_link_annotations
LaTeX	analyze_latex, prepare_latex, make_latex_accessible
Education	get_accessibility_tutorial

🏗️ Built on Standards

• MCP (Model Context Protocol) - Open standard for AI tool integration
• PDF/UA (ISO 14289) - International PDF accessibility standard
• veraPDF - Industry-standard validation engine
• WCAG 2.1 - Web Content Accessibility Guidelines alignment

---

📖 API Reference

Agent-Optimized Endpoints

These endpoints return clean, structured JSON designed for AI agent consumption:

POST /agent/validate

Validate PDF against PDF/UA with MorphMind Score.

curl -X POST http://localhost:8080/agent/validate \
  -H "Content-Type: application/json" \
  -d '{"arguments": {"pdf_path": "/path/to/file.pdf"}}'

Response:

{
  "success": true,
  "score": 82,
  "grade": "B",
  "compliant": false,
  "summary": {
    "passed_rules": 100,
    "failed_rules": 4
  },
  "issues_by_severity": {
    "critical": 0,
    "serious": 3,
    "moderate": 1,
    "minor": 0
  },
  "failures": [
    {
      "clause": "7.1",
      "test_number": "10",
      "description": "DisplayDocTitle key missing"
    }
  ]
}

POST /agent/make-accessible

Full remediation pipeline with AI alt-text generation.

curl -X POST http://localhost:8080/agent/make-accessible \
  -H "Content-Type: application/json" \
  -d '{"arguments": {"pdf_path": "file_id", "document_type": "technical report"}}'

Response:

{
  "success": true,
  "output_file": "abc123_document_accessible.pdf",
  "figures_processed": 3,
  "structure_enhancements": {
    "headings_tagged": 15,
    "links_fixed": 8,
    "xmp_metadata_added": true
  }
}

Standard Endpoints

Endpoint	Method	Description
/health	GET	Health check
/tools	GET	List all available tools
/tools/{name}	POST	Execute any tool by name
/upload	POST	Upload PDF file, returns file_id
/download/{file_id}	GET	Download processed file
/batch	POST	Execute multiple tools in sequence

---

🏛️ Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         Your Application                         │
│                  (AgentLab, MCP Clients, etc.)                  │
└───────────────────────────────┬─────────────────────────────────┘
                                │
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                    Accessibility Agent Backend                   │
│  ┌─────────────────────────┐  ┌─────────────────────────────┐  │
│  │  HTTP API (Port 8080)   │  │  MCP Transport (Port 8081)  │  │
│  │  • /agent/validate      │  │  • Streamable HTTP          │  │
│  │  • /agent/make-accessible│  │  • API Key authenticated    │  │
│  │  • /upload, /download   │  │                             │  │
│  └───────────┬─────────────┘  └───────────┬─────────────────┘  │
│              │                            │                     │
│              └──────────┬─────────────────┘                     │
│                         ▼                                       │
│              ┌─────────────────────────┐                       │
│              │   src/mcp_server.py     │                       │
│              │   25 Accessibility Tools │                       │
│              └───────────┬─────────────┘                       │
│                          │                                      │
│    ┌─────────────────────┼─────────────────────┐               │
│    ▼                     ▼                     ▼               │
│ ┌──────────┐      ┌────────────┐      ┌────────────┐          │
│ │ veraPDF  │      │  Gemini AI │      │  pikepdf   │          │
│ │Validation│      │  Alt-text  │      │PDF Editing │          │
│ └──────────┘      └────────────┘      └────────────┘          │
└─────────────────────────────────────────────────────────────────┘

---

🚢 Deployment

Self-Hosted Deployment

Deploy on any cloud provider (AWS, GCP, Azure) or on-premises:

# Clone and setup
git clone https://github.com/AIScientists-Dev/AI-Powered-PDF-Accessibility.git
cd AI-Powered-PDF-Accessibility

# Configure environment
cp .env.example .env
# Edit .env with your GEMINI_API_KEY

# Run with systemd (see deploy/ folder for service files)
./deploy/update-ec2.sh

CI/CD: Push to main triggers GitHub Actions deployment.

Environment Variables

Variable	Description	Required
GEMINI_API_KEY	Google Gemini API key for alt-text generation	Yes
PORT	HTTP server port (default: 8080)	No
HOST	Bind address (default: 0.0.0.0)	No
MCP_API_KEYS	Comma-separated API keys for MCP transport	No

---

🤝 Contributing

We welcome contributions! This project is open source under the MIT License.

Areas Where Help is Needed

• 🌍 Internationalization - Support for more languages
• 📊 Table remediation - Improve complex table handling
• 📐 Math accessibility - Better equation support
• 🧪 Testing - More diverse PDF test cases
• 📚 Documentation - Tutorials and guides

Development Setup

# Fork and clone
git clone https://github.com/YOUR_USERNAME/AI-Powered-PDF-Accessibility.git

# Install dev dependencies
pip install -r requirements.txt
pip install pytest black flake8

# Run tests
pytest

# Format code
black .

---

📊 Why Accessibility Matters

• 1 billion+ people worldwide have disabilities (WHO)
• 90%+ of PDFs lack proper accessibility features
• Legal requirements in most countries (ADA, Section 508, EU Directive)
• 2-4 hours → 2-4 minutes: Time savings with AI automation

Built-in Educational Content

Ask the agent about accessibility:

curl -X POST http://localhost:8080/tools/get_accessibility_tutorial \
  -H "Content-Type: application/json" \
  -d '{"arguments": {"topic": "what_is_accessibility"}}'

Topics available:

• what_is_accessibility - Introduction to PDF accessibility
• common_struggles - Challenges people face
• how_we_help - How AI agents solve these problems
• getting_started - Quick start guide
• about_project - About this project

---

📜 License

MIT License - Use freely, contribute back if you can.

---

🔗 Links

• Try it now: agentlab.morphmind.ai — Free trial available
• MorphMind: morphmind.ai
• MCP Specification: modelcontextprotocol.io
• veraPDF: verapdf.org
• PDF/UA Standard: ISO 14289

---

Built with ❤️ by MorphMind