Ancient Greek Reader

A reader for Ancient Greek in the browser: Homer's Iliad and Odyssey, Herodotus and Thucydides, with every word's parsing, its role in the sentence and its dictionary entry, and, in the books that have them, a gloss and a note on every word and translations of every line.

Read it online: greek.kenaldo.dev

The reader only reads saved data: it asks no model anything, needs no account, and keeps nothing about you on the server (where you stopped reading stays in your browser; the site counts visits by day, with nothing that tells one visitor from another). It works on a monitor, a laptop, a tablet and a phone. The glosses, notes and translations were written once by Claude Opus 5.5, Anthropic's model: they can be wrong, so check them against the parsing and the dictionary. The reader tags them AI, and the treebanks' parsing and LSJ's definitions Perseus, GLAUx and LSJ.

More screens: the Library, an iPad, a phone.

What each text has

Text	Parsing and roles	Glosses, notes and translations
Homer, Iliad	The Perseus treebank, annotated by hand	Every book
Homer, Odyssey	The Perseus treebank, annotated by hand	Coming
Herodotus, Histories	GLAUx: part by hand, the rest by its parser	Coming
Thucydides, History of the Peloponnesian War	GLAUx: part by hand, the rest by its parser	Coming

Every text has the Greek (Perseus), search that ignores accents and breathings, each word's parsing and role, LSJ's entry, how often its lemma occurs, and the structure of each sentence. The Library labels each text and book with what it has.

Reading

• Each book is one page you scroll through; your place follows the scroll. Go to takes Iliad 1.42, Odyssey 5, or 1.42 in the text you are reading.
• Click (or tap) a word for its meaning, parsing, role in the sentence, dictionary form, use and LSJ's entry.
• Click a line number, then Shift+click another (on a touch screen, tap the first and the last) for the passage's fluent and literal translations and its grammar, word by word or as the outline of each sentence.
• Interlinear shows a gloss under every word.

docs/reader.md describes every screen and layout.

Running it yourself

To read the texts, you don't need to: they are at greek.kenaldo.dev, with everything the reader's database holds. To run the reader on your own computer, you need Docker, or Python 3.12 or newer, and about 600 MB of disk for the data.

1. The code:

   git clone https://github.com/Kenaldo9/ancient-greek-reader.git
   cd ancient-greek-reader
   cp .env.example .env
   

2. Start it, with Docker:

   docker compose up -d          # http://localhost:8999
   

   The container runs as user and group 1000, so that files it writes into data/ stay yours; if your IDs differ (id -u, id -g), set GREEK_UID and GREEK_GID in .env. After pulling new code, run docker compose up -d --build.

   Or with Python:

   python -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt
   python greek.py web           # http://127.0.0.1:8000
   

3. The texts. The reader's own database is not offered for download for now; it may be published later, perhaps once every text has its glosses, notes and translations, but that is not decided (docs/data.md). You can build one from the open sources, free (with Docker, docker compose run --rm greek in place of python greek.py):

   python greek.py fetch Homer       # pick the Iliad; note its text ID
   python greek.py study import-treebank TEXT_ID --work iliad
   python greek.py study import-lsj  # LSJ's entries for its lemmas
   

   The Library then shows the Iliad with its parsing and dictionary. Glosses, notes and translations are made with greek study batches, which need an Anthropic API key in .env and cost money (docs/cli.md).

Newer code upgrades the database when it starts, after copying it beside itself (Database upgrades). The same program is a command line too: python greek.py --help, or docker compose run --rm greek --help (docs/cli.md). It can read and search in the terminal, fetch more texts from Perseus, and make study data (which needs an Anthropic API key in .env, and costs money).

How the study data was made

greek study sent each passage of the Iliad (whole sentences, up to ten lines) to Claude Opus 5.5 in Message Batches, with every word's lemma and parse from the treebank, and saved the answers: a gloss and a note on each word, a literal translation of each line, a fluent translation of the passage, and a remark wherever the model disagreed with the treebank. Use labels, dictionary forms, verb constructions and principal parts were made the same way. Answers are checked when they are collected, and missing ones asked again. The prompts are in greek/api/ (study.py, uses.py, forms.py and parts.py), made for each work from its profile in works.py; docs/data.md has the details.

Sources and credits

• Greek texts: the Perseus Digital Library (Tufts University), canonical-greekLit.
• Parsing and syntax of Homer: the Ancient Greek Dependency Treebank 2.1 (Perseus Digital Library).
• Parsing and syntax of Herodotus and Thucydides: the GLAUx corpus (Alek Keersmaekers, KU Leuven), with hand annotation from the PROIEL treebank (Dag Haug and others) and Vanessa Gorman's treebanks.
• Dictionary: Liddell, Scott and Jones, A Greek-English Lexicon (Perseus Digital Library, lexica).
• Glosses, notes, translations and word details: Claude Opus 5.5 (Anthropic).
• Typefaces: Gentium Book Plus (SIL International), IBM Plex Sans and IBM Plex Mono, under the SIL Open Font License (greek/web/static/fonts/).
• htmx 1.9.12, under the Zero-Clause BSD licence (htmx-LICENSE.txt), served from greek/web/static/.

Licences

• The code: MIT (LICENSE).
• The data: each source keeps its own licence: CC BY-SA for the texts, the treebanks, LSJ and the study data, and CC BY-NC-SA (non-commercial) for the hand-annotated sentences of Herodotus and Thucydides. DATA-LICENSE.md has each one, by table, and how to leave out the non-commercial part.

Mistakes and feedback

The reader is a hobby project, kept up as time allows. If a gloss, a note or a translation is wrong, please open an issue or email [email protected], with the book and line. There is no promise of a reply or of new features.

Development

ARCHITECTURE.md explains how the code is put together, and CLAUDE.md its conventions, tests and screenshot checks.

pip install httpx                       # for the route tests
python -m pytest tests/ -q              # all tests
pip install playwright && playwright install chromium webkit   # optional browser tests

Tests use their own temporary databases and never call a model. Browser tests skip when Playwright or a browser is missing.