Ragazza

The name "Ragazza" was chosen as it's memorable, simple, and cleverly contains "RAG" (Retrieval-Augmented Generation) within its spelling.

A tool to convert PDF slides into markdown format with AI-powered content analysis, suitable for loading into LLM Models.

Features

• Extracts text content from PDF slides
• Generates visual descriptions using Claude Models (or any other AWS Bedrock availables)
• Provides educational purpose analysis for each slide
• Supports error handling and retry mechanisms
• Progress tracking with tqdm
• Comprehensive logging

Installation

Using pip

pip install ragazza

From source

1. Clone the repository:

   git clone https://github.com/evereven-tech/ragazza.git
   cd ragazza

2. Use make commands for development:

   make help         # Show all available commands
   make install      # Install package in production mode
   make install-dev  # Install in development mode with dev dependencies
   make build        # Build package distribution
   make lint         # Check style with flake8
   make test         # Run tests
   make clean        # Clean up build artifacts

System Dependencies

You'll need to install Poppler for PDF processing:

# Ubuntu/Debian
sudo apt-get install poppler-utils

# macOS
brew install poppler

# Windows: Download and add Poppler to PATH

AWS Configuration

This tool requires access to AWS Bedrock to use Claude AI models for content analysis:

1. Ensure you have an AWS account with Bedrock access enabled
2. Request model access for Claude models in the AWS Bedrock console
3. Configure your AWS credentials:

   pip install awscli
   aws configure

4. Enter your AWS Access Key ID, Secret Access Key, and set default region to 'us-east-1'

Important notes:

• Using AWS Bedrock incurs costs based on token usage
• AWS Bedrock may not be available in all regions
• Your AWS user/role needs permissions for 'bedrock:InvokeModel'
• If you don't have AWS Bedrock access, this tool cannot function properly

Usage

Basic usage:

ragazza input.pdf output.md

Advanced options:

ragazza --model "anthropic.claude-3-5-sonnet-20241022-v2:0" --max-tokens 1000 input.pdf output.md

Output

The script generates:

• A markdown file with structured content for each slide
• Temporary images in ./tmp directory (automatically cleaned up)
• A log file (ragazza.log) with processing details