v1.1.0

new documentation and better logging after the release of a new ocr model

v1.0.2 - Initial Public Release

OCR Pipeline for Historical Print Periodicals using Mistral AI

Publication-ready OCR pipeline for batch processing scanned historical periodicals.

Features

• Structured OCR with Markdown formatting preserving document layout
• Automatic PDF splitting for large files (>50 MB or >1000 pages)
• SQLite checkpoint system for resumable processing after interruptions
• Multiple output formats: Markdown (.md), Plain Text (.txt), JSON metadata
• Robust error handling with exponential backoff retry logic
• Batch processing with progress tracking and cost estimation

Technical Details

• Language: Python 3.8+
• Framework: Jupyter Notebook
• OCR Engine: Mistral AI Vision API (mistral-ocr-latest)
• License: MIT
• Institution: BBF | Research Library for the History of Education in Berlin

Use Cases

Designed for researchers and digital humanities practitioners working with:

• Historical periodicals and journals
• Scanned newspaper archives
• Magazine collections
• Academic journal archives

Documentation

Complete documentation included:

• Installation and setup guide
• System architecture documentation
• Prompt engineering details
• Zenodo publication metadata

Requirements

• Mistral AI API key (registration at console.mistral.ai)
• Python 3.8 or higher
• See requirements.txt for complete dependency list

Citation

See CITATION.cff for structured citation metadata or use the Zenodo DOI (assigned upon publication).

Author: Marco Lorenz (ORCID: 0000-0002-0903-2100)

V.1.0.1 Initial Public Release

OCR Pipeline for Historical Print Periodicals using Mistral AI

Publication-ready OCR pipeline for batch processing scanned historical periodicals.

Features

• Structured OCR with Markdown formatting preserving document layout
• Automatic PDF splitting for large files (>50 MB or >1000 pages)
• SQLite checkpoint system for resumable processing after interruptions
• Multiple output formats: Markdown (.md), Plain Text (.txt), JSON metadata
• Robust error handling with exponential backoff retry logic
• Batch processing with progress tracking and cost estimation

Technical Details

• Language: Python 3.8+
• Framework: Jupyter Notebook
• OCR Engine: Mistral AI Vision API (mistral-ocr-latest)
• License: MIT
• Institution: BBF | Research Library for the History of Education in Berlin

Use Cases

Designed for researchers and digital humanities practitioners working with:

• Historical periodicals and journals
• Scanned newspaper archives
• Magazine collections
• Academic journal archives

Documentation

Complete documentation included:

• Installation and setup guide
• System architecture documentation
• Prompt engineering details
• Zenodo publication metadata

Requirements

• Mistral AI API key (registration at console.mistral.ai)
• Python 3.8 or higher
• See requirements.txt for complete dependency list

Citation

See CITATION.cff for structured citation metadata or use the Zenodo DOI (assigned upon publication).

Author: Marco Lorenz (ORCID: 0000-0002-0903-2100)

v1.0.0 - Initial Public Release

OCR Pipeline for Historical Print Periodicals using Mistral AI

Publication-ready OCR pipeline for batch processing scanned historical periodicals.

Features

• Structured OCR with Markdown formatting preserving document layout
• Automatic PDF splitting for large files (>50 MB or >1000 pages)
• SQLite checkpoint system for resumable processing after interruptions
• Multiple output formats: Markdown (.md), Plain Text (.txt), JSON metadata
• Robust error handling with exponential backoff retry logic
• Batch processing with progress tracking and cost estimation

Technical Details

• Language: Python 3.8+
• Framework: Jupyter Notebook
• OCR Engine: Mistral AI Vision API (mistral-ocr-latest)
• License: MIT
• Institution: BBF | Research Library for the History of Education in Berlin

Use Cases

Designed for researchers and digital humanities practitioners working with:

• Historical periodicals and journals
• Scanned newspaper archives
• Magazine collections
• Academic journal archives

Documentation

Complete documentation included:

• Installation and setup guide
• System architecture documentation
• Prompt engineering details
• Zenodo publication metadata

Requirements

• Mistral AI API key (registration at console.mistral.ai)
• Python 3.8 or higher
• See requirements.txt for complete dependency list

Citation

See CITATION.cff for structured citation metadata or use the Zenodo DOI (assigned upon publication).

Author: Marco Lorenz (ORCID: 0000-0002-0903-2100)