Documentation

v0.7.0Last updated March 7, 2026See changelog

Getting Started

Run the Python extraction script to convert the Voynich Manuscript PDF into individual high-res page images.

Set up the SQLite database with page metadata, seed hypotheses, and prepare the analysis pipeline.

Launch the web app to browse pages, record findings, track hypotheses, and run AI analyses.

Quick Start

cd processing && python extract_pages.py /path/to/manuscript.pdfpython init_db.pycd ../app && npm install && npm run dev

Browse all 176 high-res Yale scans with section filtering, search, and zoom controls.

Track and test theories about the manuscript's origin, language, and purpose.

Record observations per page with descriptions, types, and confidence scores.

Claude vision API analyzes page imagery — illustrations, layouts, glyph patterns.

Segment and catalog individual glyphs from manuscript pages for frequency analysis.

Statistical analysis of glyph sequences, word boundaries, and section correlations.

Endpoints for pages, images, hypotheses, findings, and analyses with field-level PATCH.

SQLite (dev) / PostgreSQL (prod) with tables for pages, hypotheses, findings, analyses, annotations.

Python scripts for PDF extraction, database initialization, and batch processing.

Python writes to SQLite + filesystem (data processing, extraction, analysis)

Next.js reads from SQLite + filesystem, writes only findings/annotations via API

Never store image blobs in SQLite — always as files

Need help or want to contribute?

This project is an experimental AI-powered analysis of the Voynich Manuscript (MS 408, Beinecke Rare Book & Manuscript Library, Yale University).