wyrdcraeft

Process Old English texts into structured JSON and generate morphology.

Why wyrdcraeft?

If you work with Old English (Anglo-Saxon) texts - editions, corpora, translation tooling, or digital humanities projects - you often need a single pipeline that turns raw or marked-up sources into a consistent, machine-readable form. wyrdcraeft provides that.

• It ingests plain text and TEI XML, converts them into a standard JSON schema that is prose, verse and dialogue aware
• Provides diacritic restoration Old English texts that have no diacritic marks.
• Includes an Old English morphology generator based on established lexical and grammatical resources.
• Provides other minor utilities for working with Old English text.

Use it from the command line or from Python, and avoid ad-hoc scripts and format fragmentation.

Features

• Ingest Old English texts from text files and TEI XML.
• Convert to a standard JSON format via deterministic heuristics or TEI parsing.
• Handle both prose and verse (paragraphs, verse lines, dialogue, sections).
• Generate Old English morphology forms using the migrated Python implementation from Ondřej Tichý's Perl-based generator (based on the Bosworth & Toller, An Anglo-Saxon Dictionary, 1898, and Wright & Wright, Old English Grammar, 1908).
• Diacritic workflows: macron restoration and disambiguation tooling for normalized forms.

Installation

Prerequisites: Python 3.11–3.13.

From PyPI with pip:

pip install wyrdcraeft
wyrdcraeft --help

With uv:

sh -c "$(curl -fsSL https://astral.sh/uv/install)"
uv tool install wyrdcraeft
wyrdcraeft --help

With pipx:

pipx install wyrdcraeft
wyrdcraeft --help

From source (development):

git clone https://github.com/cmalek/wyrdcraeft.git
cd wyrdcraeft
uv sync --dev

Canonical database

Morphology and dictionary data live in one SQLite file: wyrdcraeft.sqlite3 under the OS application-data directory (override with WYRDCRAEFT_APP_DATA_DIR or app_data_dir in .wyrdcraeft.toml).

On first run (or after a schema upgrade), wyrdcraeft checks the database, applies Alembic migrations, and keeps a backup copy. If an older morphology.sqlite3 is found, it is backed up and replaced with a fresh canonical database; the CLI then prints a rebuild recipe.

Rebuild from scratch:

wyrdcraeft dictionary build --with-morphology

Use --source PATH to index a custom Bosworth-Toller file instead of the packaged default (wyrdcraeft/etc/dictionary/oe_bt.txt).

Documentation

Full documentation (installation, quickstart, CLI, Python client, configuration, FAQ): https://wyrdcraeft.readthedocs.io

For wyrdcraeft dictionary browse, the search field accepts direct Old English characters including æ Æ ð Ð þ Þ ā Ā ē Ē ī Ī ō Ō ū Ū ȳ Ȳ ǣ Ǣ ċ Ċ ġ Ġ. On macOS with the ABC Extended keyboard layout, those keys are supported directly in the browse input; on other terminals or keyboard layouts, direct typing may also work, and the on-screen character buttons remain available as a fallback.

Contributing, Licensing and Provenance

Contributing

Contributing and coding standards are described in the documentation (runbook).

Licensing and Provenance

Bosworth-Toller Old English Dictionary

The OCR extracted text of the Bosworth-Toller Old English Dictionary used in this project is from the Germanic Lexicon Project. The scanning was done by Jason Burton, B. Dan Fairchild, Margaret Hoyt, Grace Mrowicki, Michael O'Keefe, Sarah Hartman, Finlay Logan, Sean Crist, Thomas McFadden, David Harrison, and Sean Crist; that data is in the public domain.

Morphological Analyser of Old English

• The Old English morphology generator in wyrdcraeft is based on the work of Ondřej Tichý's thesis, Morphological Analyser of Old English (2017).
• The upstream morphological generator Perl code and data is (c) Ondřej Tichý, is released under the CC BY 4.0 license. The modified Perl code itself, with Madeleine Thompson's changes, can be found at github:madeleineth/tichy_oe_generator.
• Changes made to the morphology generator in this repository by the maintainers of wyrdcraeft are released under the MIT license.

All other code

• All other code implemented directly by this project's maintainers are also released under the MIT license.