VineVerse

Changelog

Generated from the knowledge base’s own log file, so it records what was actually built rather than what someone remembered to write down.

Log

2026-08-20

  • Creation: Generated the bundle from 25 upstream datasets — 12,761 documents, 55.2 MB.
  • Creation: 1189 chapters and 66 books from the Berean Standard Bible, all 31,102 verse slots preserved.
  • Structure: how the text is set, recovered from the USFM edition of the same translation — paragraphs, poetry lineation, 1189 chapters' worth of section headings, translator notes, and the words of Christ marked at the phrase rather than the whole verse. Recorded as word offsets into the plain text in x-bk.structure, so the quoted wording stays byte-identical to its source; 22,957 of 23,013 positioned verses were confirmed to agree across both editions before any offset was written.
  • Structure: red letter now comes from the edition's own \wj marking rather than from the speaker dataset, which records whole verses and so coloured "Jesus replied," along with the reply. x-bk.speakers is kept — it names every speaker, which the edition's marking cannot. The two accounts do not agree everywhere and neither is simply wrong; npx tsx scripts/dev/check-redletter.ts --verses lists every verse where they differ.
  • Creation: 36,803 word-level mentions — every name in the text that points at a person or place document, carried on the relation as the exact word rather than as markup in the scripture, so the Berean text in this bundle still matches its source byte for byte. 35,534 attested in the verse, 1,269 carried from elsewhere in the chapter. 1,047 attestations left unmarked because the source records them as naming the group, not the individual. 221 carried matches declined for being ordinary English words, 267 declined for naming a tribe or territory rather than its patriarch. 68 surfaces refused as ambiguous within their chapter (commonest: "Nazarene" in 13, "Babylon" in 10, "Ammon" in 9, "Tubal" in 9, "Aram" in 8, "Jair" in 8).
  • Creation: 3009 people with 5,450 recorded relationships.
  • Creation: 1342 places, of which 777 carry competing modern identifications.
  • Creation: 4,133 person-to-place links joining 1,080 people to 773 places — 31 birthplaces, 15 places of death and 56 other event locations from BibleData's event table, the rest same-verse co-occurrence typed co-occurs-with so it cannot be mistaken for a claim of presence.
  • Creation: 450 events under the share-alike partition at /sa/events/, joined to the corpus by 1,066 chapter links, 808 participants across 428 events and 474 locations across 305. Those links point out of the share-alike partition and never back into it, so the permissive tree is unchanged.
  • Creation: 3,005 themes from Nave's Topical Bible, joined to the corpus by 40,878 develops edges across 1,180 chapters. A further 1,002 topical articles name exactly one person and 660 name exactly one place; those appear on the existing document rather than as a second node with the same name.
  • Creation: 613 commandments of the traditional enumeration, each with its verse in English, Hebrew and Greek and its place in the Sefer HaChinuch and Mishneh Torah.
  • Creation: original-language names throughout — 3,747 person labels and 125 place labels carrying Hebrew and Greek forms, transliterations, meanings and Strong's numbers. Those labels cite 1,856 distinct Hebrew words, which with their root closure make 2,901 lexeme documents. Greek numbers are recorded but link nowhere: this bundle carries no Greek lexicon, and a link to a missing document would be worse than an honest dead number.
  • Creation: 165 epochs — lifetimes, reigns, judgeships and ministries dated from Creation. This is the first chronology in the permissive tree; the event graph remains share-alike and quarantined.
  • Creation: source documents record the books each dataset covers, so every type: Source document is part of the graph rather than a footnote beside it.

Known gaps

  • 245 Theographic people and 43 places match nothing in the permissive datasets, so events mentioning them carry fewer links. The largest are "Azariah", "Elnathan", "Abinadab", "Adaiah" — the permissive people dataset does not carry God, Israel or the Holy Spirit as named individuals at all. Matching them on shared verses alone would also make Aher into Hushim, because two people named once in the same genealogical verse have identical verse sets; the join refuses ties rather than guessing.
  • Chapter descriptions are not yet written. They are the one part of this bundle that needs a language model, and nothing has been generated yet — no document claims a description it does not have.
  • 6 event locations resolve to no mapped place — "West of Eden", "Cave near Zoar", "Cave of Machpelah" and others are descriptive phrases rather than names the geocoding data carries. Pointing them at the nearest named site would fabricate a precision the source does not have, so those links are absent.
  • 56 verses carry no structure. The plain text and the USFM edition of the same translation disagree about their wording — a space before a comma, a census total written ". . . 40,500", a stray "vvv" in Genesis 35:18 — and a word offset is meaningless against text it was not computed from. Those verses read as plain paragraphs. Named in full: 1Chr.11.5, 1Chr.11.18, 1Chr.17.10, 1Chr.21.4, 1Chr.25.3, 1Kgs.4.15, 1Kgs.14.14, 1Kgs.20.34, 1Sam.1.20, 1Sam.8.19, 2Chr.18.6, 2Chr.23.18, 2Chr.34.18, 2Kgs.22.10, 2Sam.5.7, 2Sam.23.16, 2Sam.24.4, Acts.1.15, Acts.4.36, Acts.11.28, Acts.13.1, Exod.38.28, Ezek.14.22, Gen.35.18, Hab.3.19, Heb.7.11, Judg.8.26, Judg.13.11, Judg.16.14, Jer.22.24, Jer.29.19, Jer.48.14, John.6.1, John.18.40, John.20.16, John.21.7, Job.6.23, Job.8.18, Lev.7.9, Luke.9.33, Luke.19.26, Matt.1.23, Mark.7.11, Mark.15.42, Nah.2.10, Num.6.23, Num.26.18, Num.26.22, Num.26.25, Num.26.27, Num.26.34, Num.26.37, Num.26.41, Num.26.47, Num.26.50, Ps.66.10.
  • No document has been reviewed by a person. Every verified entry is a mechanical check, so every trust tier in this bundle is machine-confirmed at best.