pdf2html-ai 0.1.3

AI-powered PDF to HTML conversion using mistral-ocr and pixtral-12b

PDF2HTML AI

A Python package for converting PDF documents to accessible HTML using Mistral OCR and Pixtral 12B. This tool processes PDFs and generates WCAG-compliant HTML output with enhanced accessibility features.

Features

• Process local PDF files or download from URLs
• OCR processing with Mistral OCR
• Generate accessible alt text for images using Pixtral 12B
• Convert to WCAG-compliant accessible HTML
• Enhance tables with proper accessibility features
• Save output as HTML file
• Option to open result in browser

Requirements

• Python 3.10+
• Mistral API key
• Required Python packages (automatically installed with pip):
  • mistralai
  • requests
  • python-dotenv
  • PyPDF2

Installation

Option 1: Install from PyPI (Recommended)

pip install pdf2html-ai

Option 2: Install from Repository

1. Clone the repository:

   git clone https://github.com/mystique920/ai-powered-pdf2html
   cd ai-powered-pdf2html
   

2. Install required packages:

   pip install -r requirements.txt
   

Quick Start

After installing the package, you can use it in your Python code or via the command line.

Using as a Python Package

from pdf2html_ai import process_pdf_with_ocr, convert_ocr_to_accessible_html
from mistralai import Mistral

# Initialize Mistral client
client = Mistral(api_key="your_api_key_here")

# Process a local PDF file
with open("document.pdf", "rb") as f:
    file_content = f.read()
    
ocr_result = process_pdf_with_ocr(client, file_content, "document.pdf")
html_content = convert_ocr_to_accessible_html(client, ocr_result)

# Save the HTML output
with open("output.html", "w", encoding="utf-8") as f:
    f.write(html_content)

Using the Command Line

After installing the package, you can use the pdf2html-ai command directly:

Process a local PDF file:

pdf2html --file path/to/your/document.pdf

Process a PDF from a URL:

pdf2html --url https://example.com/document.pdf

Alternatively, you can still use the module directly:

python -m pdf2html_ai.processor --file path/to/your/document.pdf

Example Scripts

Several example scripts are provided to help you get started:

• examples/example.py - An interactive example that guides you through the options
• tests/test_local_pdf.py - A test script for processing a local PDF file
• tests/test_url_pdf.py - A test script for processing a PDF from a URL

To run the interactive example:

python examples/example.py

Command-line Options

• --file, -f: Path to local PDF file
• --url, -u: URL to PDF file
• --api-key, -k: Mistral API key
• --output, -o: Output HTML file path (default: output.html)
• --max-images, -m: Maximum number of images to process (default: all)
• --open-browser, -b: Open the output HTML in browser after processing

Examples

Process a local file with a custom API key and open in browser:

pdf2html --file document.pdf --api-key YOUR_API_KEY --open-browser

Process a PDF from URL and save to a custom output file:

pdf2html --url https://example.com/document.pdf --output result.html

Process a file but limit image processing to 5 images:

pdf2html --file document.pdf --max-images 5

API Key Setup

The script requires a valid Mistral API key to function. There are two ways to provide the API key:

1. Create a .env file in your project directory with the following content:

   MISTRAL_API_KEY=your_api_key_here
   

2. Provide the API key directly using the --api-key command-line argument:

   pdf2html-ai --file document.pdf --api-key your_api_key_here
   

Notes

• Processing large PDFs may take some time
• Image alt text generation uses the Pixtral 12B model
• The HTML output is designed to be WCAG-compliant for accessibility
• You can limit the number of images processed to save API usage
• The code for this application is based on a public Google Colab notebook
• The original code is from this repository: https://github.com/coldplazma/Accessible-OCR-Mistral-
• This tool was mostly modified and extended using AI tools. Use this tool at your own risk