v1.3.0 — from Sumer to the scriptoria

26 July 2026 · Nabu news

Version 1.3.0 is tagged — the largest span between any two tags of this library: from 801 thousand documents / 28.2 million passages at v1.2.0 (22 July) to 947,273 documents / 66,755,518 passages across 96 sources four days later.

The Arabic phase more than doubled the corpus in a day: OpenITI, the premodern Islamicate library whole — hadith, history, falsafa, the dīwāns — with its Persian shelf, making Classical Arabic the largest language on the shelves (33.3 million passages).

The instruments phase gave verse a meter layer (167,732 scanned lines; search --meter H filters to hexameters and Iliad 1.1 knows what it is), moved the Pleiades gazetteer into a derived index (place lookups from seconds to milliseconds), and finished the MCP surface’s user pass — eleven read-only tools.

The Romance phase opened the twenty-first desk: the MGH critical editions, digilibLT’s late-antique prose, the Base de français médiéval from the Serments de Strasbourg onward, the UD Romance treebanks — four new gold languages in a day, CroALa dual-tagged as the continuum’s Latin side.

The Rabbinic and Geʿez phase closed the span: the Mishnah in three complete Hebrew editions, both Talmuds with daf-grain citations, the Tosefta; the first Ethiopic shelf — 1 Enoch and Jubilees, complete only in Geʿez, with Dillmann’s lexicon and the TraCES gold corpus — under the twenty-second desk; the inscriptions of Italy (EDR, 115,590 records); Urartian, gold-lemmatized; Middle Low German; and the comparativist loanword layer (WOLD, CLICS³, the StarLing Kartvelian base).

The census at the tag, all live-read on 26 July 2026: 19,223,970 gold lemma rows in thirty-seven languages (28.1 million more at a labelled silver tier), 1,391,759 dictionary entries on one hundred shelves, 549,771 dated-and-placed documents on the timeline, and a written dossier behind every one of the 305 held language codes.

As ever: local, license-honest per document, rebuildable from canonical bytes, and readable — never writable — by the AI tools at the researcher’s desk.

← All news  ·  Atom feed