This page renders CANARY.md, the release-notes file kept in the museum's public repository. Every change lands with its entry in the same commit, grouped by kind; a rollback is recorded, never erased. The entries are kept in English as the engineering record.
Current version: v2.27.0 · 2026-11-21
v2.27.0 — 2026-11-21
Added
- Eleven new records from pages 121 to 180 of the Arabic translation of Rashid al-Din's Jami' al-Tawarikh (volume 2, part 1): later historians and works that continue or quote it (Mirkhwand, Khwandamir, Haydar al-Razi, al-Banakati, Ali Yazdi, Tarikh al-Khata, Tawdihat, Miftah al-Tafasir, Sultaniyya, Tawa'if al-Haqa'iq) and the place Khansa, each citing the book's pages, with uncertain readings flagged.
- Eight new records from pages 61 to 120 of the Arabic translation of Rashid al-Din's Jami' al-Tawarikh (volume 2, part 1): Ghiyath al-Din ibn Rashid, Ibrahim ibn Rashid al-Din, Narin Taghay, Izz al-Din Talib Dalqandi; the 717 dismissal and the 718 and 736 executions of the Rashid family and Ghiyath; and the 710 completion of the book, each giving the book's account, attributed to the book, with a note where no independent Sunni view is sourced.
- Seventeen new records from the first 60 scanned pages of the Arabic translation of Rashid al-Din's Jami' al-Tawarikh (volume 2, part 1: the Ilkhanid history): the book and Wassaf's history as sources; the men around Ghazan's court, including Sadr Jahan, Sa'd al-Din al-Sawaji, Ali Shah al-Jilani, Wassaf, Hafiz-i Abru and the Alid naqib Taj al-Din; the events of 697, 700, 702, 706 and 710 AH; the Rab' al-Rashidi quarter at Tabriz; and two open questions, on Rashid's birth year and on the claim of his Jewish origin, each giving the book's accounts side by side with a note where no Sunni view is sourced.
- A new hall: the Arabic Language Browser («متصفح اللغة العربية لغة القرآن»), at `/fr/arabiyya/` in Arabic, English, Urdu, French and Spanish. It opens on the ring of the letters, the Arabic letters in the order al-Khalil ibn Ahmad gave them by place of articulation. Below it stand the fifteen verses in which the Qur'an speaks of its own tongue, nine chapters of records (Arabic as the language of the Qur'an, the inimitability of its language, origins and history, letters and script, the sciences of Arabic, dialects, Arabic in the world, the scholars, the books), and 77 passages from the hadith, the seerah, the tafsir and the other harvested books that speak about Arabic. The passages are found by their wording each time the site is built, so a book harvested later joins on its own; they are not reviewed one by one. The home page and the Explore menu carry the new entry. The dataset is at
/api/v1/arabiyya.json. - 22 new records with full prose in five languages, among them: what four classical commentators say on why the Qur'an is Arabic; the men of Quraysh who heard it and what they said; the challenge verses and the books on inimitability (al-Rummani, al-Khattabi, al-Baqillani, al-Jurjani); the reports on learning Arabic with each one's grading; the debate on foreign words in the Qur'an; the origin of language; grammar, morphology and prosody; the dialects of the tribes and today's spoken Arabic; the Semitic family; the Arab countries; Arabic at the United Nations; and what Arabic gave other languages. Every judgement is given under the name of the scholar who made it. The prose and its translations were written by an AI model from the cited sources and have not been read by a human.
Data
- 92 records the museum already held (grammarians, lexicographers, books, scripts, the Arabs before Islam) now carry the browser's tags; nothing else in them changed. 42 source records were added for the editions and pages read on 21 November 2026.
Tests
scripts/arabiyya.test.tschecks the detector: the verses, the known false positives (a man named Ibn al-Arabi, a bare citation of the dictionary Lisan al-Arab), and that a passage from a book harvested later is picked up.
Data (Arabic Language Browser, books chapter)
- Five records for the classic books on the Arabic language joined the Arabic Language Browser's "books" chapter, each with full prose in Arabic, English, Urdu, French and Spanish and cited page by page: al-Khasa'is (Ibn Jinni), al-Muzhir fi Ulum al-Lugha (al-Suyuti), al-Sahibi fi Fiqh al-Lugha (Ibn Faris), Dala'il al-I'jaz (Abd al-Qahir al-Jurjani) and Mughni al-Labib an Kutub al-A'arib (Ibn Hisham al-Ansari). A reader of the Arabiyya chapters can now open the book behind each quotation, see what part of it was read, and follow links to the grammar, origin-of-language and inimitability pages. The prose says plainly which chapters were read and which were not.
- One new source record, the Damascus (Dar al-Fikr, 1985) edition of Mughni al-Labib as shown on Shamela, with its page-numbering difficulty stated on the record. The page that closes the author's preface shows the number 1316 between pages 15 and 17; it is cited as page 16 by position and marked as an inference.
- Two difficulties are stated on the records rather than hidden: the Shamela copy of al-Sahibi carries an introduction signed by Ahmad Hasan Basaj while the museum's source record names the publisher Muhammad Ali Baydun; and the Shamela text of Dala'il al-I'jaz gives a reference to Hud 13 beside a verse whose words are those of al-Isra 88.
- Fixed: three Arabic Language Browser concept records carried the wings value
places, which the schema does not allow; it is nowplace.
Fixed
- The Arabic references page is back under its page-weight budget. The landing of
/fr/references/printed all 576 web-page records, and at 567 KB it was over the 500 KB listing budget, which failed the whole-site check of build 44988045069. The landing now shows the first 150 web pages, as it already did for the books, with a link to the full list on/fr/references/web/and its numbered sub-pages. Every record is still listed, and the other lists on the landing are unchanged. - The Khansa place record passes validation again. The record added in the Jami' al-Tawarikh batch carried a
namesfield that the place schema does not define, which failednpm run validateand two tests onmain. The name is already the record's English label, so the field is removed and nothing else in the record changes.
Removed
- A stray
wt3folder link, committed by mistake with no submodule entry, is removed from the repository. Nothing on the site changes; it stops the git warning that every build job printed while cleaning up.
Changed
- The project status page's Tranco ranking log is current to 21 November 2026. islam.museum stands at #716,997 on the Tranco list of 21 November 2026 (list 46YLX, read from the published list file of 4,338,665 domains) — up 140,948 places in a day, and 3,892,970 places above its first reading of #4,609,967 on 23 October 2026. The log's rows for 18–21 November and its Latest note record the climb.
Data (Arabic Language Browser: the sciences of rhetoric and the Arabic dialects)
- Ten new concept records in five languages, tagged for the Arabic Language Browser. Rhetoric (arabiyya-sciences): the science of rhetoric, ma'ani, bayan and badi', each with prose built from pages actually read: al-Qazwini's al-Idah (with Khafaji's commentary), al-Sakkaki's Miftah al-Ulum, Ibn al-Mu'tazz's al-Badi', Ibn Khaldun's Muqaddima and al-Jurjani's Dala'il, with Qur'anic examples pointed to by verse page and explained from the tafsir of al-Qurtubi, al-Baghawi and Ibn Kathir (no verse text typed). Dialects (arabiyya-dialects): the Gulf, the Hejaz, the Levant, Egypt, the Maghrib and Yemen, each stating plainly what the sources read say and what they do not (no speaker counts or modern features were found for most of them, and the pages say so instead of filling the gap).
- Ten new source records: al-Idah (Khafaji ed.), Miftah al-Ulum (Zarzur ed.), al-Badi' (Ibn al-Mu'tazz, Dar al-Jil) and Encyclopaedia Britannica pages on Egypt (languages), Yemen, Qatar, Bahrain, Morocco, the Maghrib, Mauritania and Saudi Arabia. The Shamela copy of al-Idah read carries only the ma'ani section, which each page notes.
- All new prose and translations are AI-written from the cited sources and have not been read by a human; Urdu, French and Spanish are machine translations.
Data
- Arabic by country, first group of eleven: eleven new concept records, "Arabic in" Saudi Arabia, Yemen, Oman, the United Arab Emirates, Qatar, Bahrain, Kuwait, Iraq, Syria, Lebanon and Jordan (al-arabiyya-fi-<country>), each in Arabic, English, Urdu, French and Spanish and tagged for the Arabic browser's world view. Each sets out the official status of Arabic from the text of the country's constitution, the spoken varieties and how the dialect groups are classified (University of Manchester's geographical groups beside Glottolog's separately named varieties, with the differences stated, not settled), the other languages the sources name, and how the standard and the spoken varieties sit together. Where a source gave no speaker count, dialect description or list of other languages, the record says so. Bahrain has no place record in the graph, so it is mentioned without a link.
- Twenty-four new source records for the pages read: eleven constitution texts from the Constitute Project, nine Library of Congress Country Studies pages (old, undated handbooks, flagged as such), two University of Manchester dialect pages and Glottolog's classification tree for Arabic. Not reached: the CIA World Factbook country pages (the publication was retired, notice of 4 February 2026), Britannica country pages (refused automated fetches), Ethnologue (refused).
Added (Arabic language browser, the Arab world chapter, countries group B)
- Eleven new concept records, "Arabic in Palestine, Egypt, Sudan, Libya, Tunisia, Algeria, Morocco, Mauritania, Somalia, Djibouti and Comoros" (
concept/al-arabiyya-fi-<country>), each taggedarabiyya-browserandarabiyya-worldand written in Arabic, English, Urdu, French and Spanish. Each sets out only what the pages read say about the place of Arabic in that state: official status, the spoken varieties as the sources name them, the other languages, and notable history of the language there. Where the Encyclopaedia Britannica and the CIA World Factbook differ (the share of Tamazight speakers in Morocco, whether French is a lingua franca in Algeria, the share of French speakers in Tunisia, the names of Fula and Comorian) both are stated and neither is chosen; each page ends with what was not read (constitutions, Ethnologue and Glottolog classifications, speaker counts). - Twenty-four new source records: eleven Britannica pages (Palestine; Sudan and three of its sub-pages; Libya People; Tunisia Languages; Algeria Languages; Somalia; Djibouti; Comoros), the Palestinian Basic Law as shown by the Constitute Project, and twelve Factbook country pages. The Factbook pages are cited from the copies archived by the Internet Archive (dates recorded on each record) because cia.gov no longer serves them. Djibouti has no place record, so its page is unlinked until one exists.
- Not read and said so on the records: the University of Manchester dialects site (its certificate was invalid at the time), Ethnologue, Glottolog, and any national constitution other than Palestine's.
Changed (Arabic Language Browser)
- Twelve passages that only mentioned the Arabic language in passing (a quoted verse inside another argument, a cited book title, a bare phrase on a dictionary entry) were taken out of the detected list after a page-by-page read, so the browser shows 65 detected passages instead of 77. Borderline ones were kept.
Data (Arabic Language Browser, dialect groups)
- The six dialect pages (the Gulf, the Hejaz, the Levant, Egypt, the Maghrib, Yemen) now carry sections on classification, main features (sounds, word forms, sentence structure, vocabulary) and speaker numbers, in Arabic, English, Urdu, French and Spanish. Classification is attributed to Glottolog, Ethnologue's free pages, Brill's Encyclopedia of Arabic Language and Linguistics and the Encyclopaedia Britannica, side by side where they differ. Speaker figures are given exactly as the research summary gave them and dated: over 30 million for Levantine Arabic (CUNY, 2024), about 8.5 million for Palestinian Arabic (Brill, 2019), 77,436,230 for Egyptian Arabic (Britannica, 4 August 2023), 24 million for Moroccan Arabic in Morocco (Yale paper of 2017 citing the 2004 census, set beside Britannica's earlier two-thirds), the 1995-1996 Yemen figures from a Stanford course archive page flagged as old, and Britannica's "more than nine-tenths of Yemenis". For the Gulf and the Hejaz the pages say plainly that no speaker count was found.
- Where the new sections come from is stated on every page. They rest on a research summary supplied by the founder, compiled with an AI research assistant, which the museum used on the founder's approval of 21 November 2026 without opening the pages it cites. 57 new source records name those pages with their exact addresses, publishers and dates, and each says it was not opened by the museum. The remarks on the older sections that no description or count was found now say they refer to the references the museum itself read.
Data and content: constitutional text and language shares on the 22 "Arabic in ..." pages (21 November 2026)
- Constitutional text section on all 22 country pages (Arabic, English, Urdu, French, Spanish). Each page now gives the constitutional document, the article and the Arabic wording as a short cited quotation, with the museum's own translation into the page's language. The wording and the official addresses come from a research summary supplied to the museum, which the founder approved using without re-checking on 21 November 2026; the museum did not open the pages the summary names, and every page and source record says so. Where the summary says no official Arabic text is published (Comoros, Somalia) the original French or Somali wording is quoted.
- Corrections to earlier statements. Saudi Arabia: the earlier page said no clause names a language; the summary gives Article 1 in Arabic as naming Arabic as the Kingdom's language (only that clause is quoted), and the page now states that the two readings disagree and have not been reconciled. Syria: the 13 March 2025 Constitutional Declaration, Article 4, replaces the 2012 text as the operative reference, which stays cited as an earlier text. Sudan: Article 6(b), added in February 2025, makes Arabic and English the official languages for the work of government and calls Arabic the most widespread national language; it replaces the earlier note that an encyclopedia footnote was unclear. Libya (2011 Declaration, Article 1), Iraq (Article 4, with Turkmen and Syriac in the administrative units where their speakers are dense), Palestine (Basic Law, Article 4, paragraph 3, beside the earlier citation), Somalia (Article 5, Somali official, Arabic the second language), Djibouti (Article 1 co-official Arabic and French, Article 93 French text authoritative) and Comoros (Article 9, French original only) are added with their own notes.
- Language shares section on all 22 pages. Each page now gives the figure the research summary lists, with what is measured, the issuing body and the year, or the sentence "No official figure" with the reason the summary gives. Figures: Djibouti (INSTAD RGPH-3, 2024: Somali 78.4, French 41.6, Arabic 26.6, Afar 25.8, multiple responses; its Arabic figure combines French and Arabic 21.6 with Arabic only 5.1), Morocco (HCP 2024, non-exclusive, and the 2014 split of Amazigh), Tunisia (INS 2024, languages read and written among the literate aged 10 and over), Iraq (Ministry of Education to UNESCO, 2020, Arabic about 85, Kurdish about 21, other 3, which sum to 109 percent in the source, and the page says so) and Algeria (no official figure; a UNESCO GEMR paper cites 20 to 30 percent Tamazight as its author's estimate). Comoros and Mauritania are given only as raw unweighted microdata cases, with the warning, quoted from the summary, that the catalog pages themselves say such counts are not official population statistics.
- Djibouti address unverified. The summary gives the INSTAD report through a google.com search link; the underlying address is recorded exactly as it appears inside that link (with doubly encoded characters) and the source record says the museum has not verified that it works.
- New source records (59). One record per official document or address for the constitutional wording, and one per address for the language shares, each with the note "wording supplied in a research file, page not opened by the museum". The pages' "what this page has not reached" sections gain a dated note that the "not read" statements describe the museum's own earlier reading and were left as written.
- content: New concept page "Is Arabic one language or several? How the varieties are classified" (concept/tasnif-al-arabiyya-lughatan-am-lahajat), in Arabic, English, Urdu, French and Spanish. It sets side by side how the Library of Congress (ISO 639-2: one language, ara), ISO 639-3 (macrolanguage ara with 28 active individual languages, two deprecated codes) and Ethnologue 29th edition 2026 (each variety its own language; 33 in its Arabic subgroup) classify Arabic, with a list of the varieties and their codes, and prefers none. Judeo-Arabic varieties sit under jrb, Hassaniyya and Maltese are outside ara in Ethnologue, Juba Arabic is a creole. Source: a research summary supplied by the founder, used with his approval of 21 Nov 2026 without the museum opening the pages; no speaker numbers are given because Ethnologue withholds them. Linked to the Standard Arabic and dialects page, the tribal-dialects page and the six regional dialect pages. Tagged arabiyya-browser, arabiyya-dialects.
- data: Three new source records (loc-iso-639-2-ara, iso-639-3-ara, ethnologue-ara) beside the existing ethnologue-arabic-subgroup, each stating that the museum did not open the page.
Data (kind: content correction, five languages)
- Classical Arabic philology: five open questions on the Arabic-browser pages are answered from a research summary supplied to the museum, which the founder authorised the museum to use without re-verifying (21 November 2026). The museum did not open the pages the summary names; each addition says so, names the pages the summary gave, and states that the founder approved its use. Eight new source records (
islamweb-siyar-makki-ibn-abi-talib,islamweb-siyar-al-hawfi,islamweb-fatwa-136136,alukah-majles-138809,usul-ai-sahibi-introduction,al-ilmiyah-sahibi-9782745157621,ablibrary-dalail-shakir-index-p634,islamweb-jurjani-risala-shafiya-sarfa) record those pages as supplied, not opened. - Mughni al-Labib (work page and source record): the page that ends the preface is page 16; the 1316 shown in the Shamela copy is a digitisation typo (the summary's answer, replacing the earlier "inference from position"). The preface now links the two grammarians it names, Makki ibn Abi Talib and al-Hawfi, and the page gives their full names and death dates from al-Dhahabi's Siyar A'lam al-Nubala' (Makki, Muharram 437 AH; al-Hawfi, 430 AH). Both person pages gain a section with the full name, the Siyar citation and the link back to the Mughni preface.
- Dala'il al-I'jaz (work page and source record): the earlier unexplained mismatch ("Hud 13" beside the verse on humans and jinn gathering, al-Isra 88) now carries an attributed explanation: in Shakir's own index Hud 13 belongs to the verse on ten invented suras and al-Isra 88 to the other verse, so the Hud 13 label is a fault of a secondary reprint or a transcription, not Shakir's reference. Not checked by the museum.
- al-Sahibi fi Fiqh al-Lugha (work page and source record): the editor's introduction is signed by Ahmad Hasan Basaj, Beirut, 21 Sha'ban 1417 AH / 1 January 1997; Muhammad Ali Baydun is the publisher. The source id keeps the publisher's name and is not renamed; a note records who the editor is.
- Learning Arabic in the reports of the Companions and Successors: a new section on the saying "Learn Arabic, for it is part of your religion" gives, from the summary, the attribution to Umar ibn al-Khattab as mawquf, Ibn Taymiyya's quotation in Iqtida' al-Sirat al-Mustaqim (1/528), the early-collection wording and chains (Ibn Abi Shayba no. 31918, Sa'id ibn Mansur, al-Bayhaqi al-Sunan al-Kubra 6/209, al-Bayhaqi Shu'ab al-Iman no. 1675), and the gradings of Ibn Hajar (Fath al-Bari 12/5) and al-Batili, set beside the editor's "sahih" already on the page, with no verdict. It replaces the earlier "full chain not found" statement and keeps saying that the museum did not read the chain itself.
- All five languages (Arabic, English, Urdu, French, Spanish) in every changed record; French and Spanish from the English; machine translation by the same model, no human read. No honorific is typed; Arabic parentheses are isolated.
Added
- Place records for two Arab states, linked from the Arabic browser: Bahrain (place/bahrain, Wikidata Q398) and Djibouti (place/djibouti, Wikidata Q977), each in Arabic, English, Urdu, French and Spanish, with names, coordinates, status and independence dates, memberships and the language situation. The facts come from a research summary of the English Wikipedia article and the Wikidata item that was supplied to the museum and approved by the founder on 21 November 2026; the museum did not open the pages, and the revision ids and the date checked that the authenticity ledger asks for are not recorded, which each record says on its face. Four source records were added for the pages (source/wikipedia-bahrain-en, source/wikidata-bahrain, source/wikipedia-djibouti-en, source/wikidata-djibouti).
Changed
- "Arabic in Bahrain" and "Arabic in Djibouti" now link to their place records in all five languages, in the prose and in the relations, replacing the statement that the museum had no place record for the state.
Content
- "What Arabic gave other languages" (concept/athar-al-arabiyya-fi-al-lughat) now carries ten example loanwords and a reported share of vocabulary for each of seven languages: Spanish, Persian, Turkish, Urdu, Swahili, Malay and Indonesian, and Hausa. Each word is given with its Arabic source word and meaning, in all five site languages. This replaces the page's earlier statement that no example words had been found. The examples and figures come from a research summary supplied to the museum; the museum did not open the pages it names, and the founder approved using it without re-verification on 21 November 2026. The page keeps the summary's hedges and shows conflicting figures side by side (Swahili: about 40% against 16 to 20% in some historical sources), and says plainly that the shares are measured in different ways and are not comparable. Words marked in the summary as coming through Persian or Omani Arabic, or with a literal sense, keep those marks.
- Fourteen new source records back the section: seven for the pages the figures come from (Wikipedia articles, Encyclopaedia Iranica, two Encyclopedia.com entries; the Wikipedia ones are marked as secondary summaries) and one Wiktionary record per language listing its ten entry pages.
Docs (Hudhud's Story Time books)
- A check and a harvest queue for the books named in the kids' story series. Of the 36 books and works the 40 scripts name or describe, 19 are already source records and 3 of those are not yet cited by their author's record; 14 are not in the museum and are queued, and 1 (Jami' al-Tawarikh, volume 2) is already downloaded and OCR'd but has no source record in
docs/delegation/harvest-queue-hudhud-books.md, and 2 (Aristotle's *Metaphysics* and Ibn Rushd's commentaries as a body) wait for the founder's decision. No source, content or entity record was changed. Twelve of the queued books were then downloaded to the microSD and entered in the harvest register; all of them, plus al-Kindi's cryptanalysis volumes and a Latin Algorithmi, were then OCRed on the founder's Mac with Apple Vision (OCR text stays on the microSD, not reviewed yet).
Data
- Arabic language browser: five records of the Arabic language academies, each with full prose in Arabic, English, Urdu, French and Spanish, tagged arabiyya-browser and arabiyya-scholars: the Arab Academy of Damascus (origin 1918-1919, Muhammad Kurd Ali president 1919-1953), the Academy of the Arabic Language in Cairo (decree 31 December 1932, inaugurated 30 January 1934), the Iraqi Academy of Sciences (1947, with Kurdish and Syriac departments since 1963), the Jordan Academy of Arabic (first 1923-1924, re-established 1 July 1976) and the Union of Arab Language Academies (13 March 1971). Each links the persons and works the museum already holds (Kurd Ali, Taha Hussein, al-Jawari, Fuad I, Ahmad Rida, Matn al-Lugha, al-Mu'jam al-Wasit, al-Mu'jam al-Kabir) and says what its sources leave unknown (journal names, exact founding days, later history). Why it matters: readers of the browser can now see who made the modern terminology and Arabisation institutions of Arabic, in their own five languages.
- New concept concept/yawm-al-lugha-al-arabiyya-al-alami (World Arabic Language Day), tagged arabiyya-browser and arabiyya-world, in five languages: the 2010 United Nations decision to give a day to each of its six official languages, the choice of 18 December for Resolution 3190 (XXVIII) of 18 December 1973, and the observance since 2012. It links, and does not repeat, the existing speaker-figures page.
- Twelve new source records (English and Arabic Wikipedia, Encyclopedia.com and the official websites of the Cairo academy, the Jordan academy and the Union). All facts come from a research summary supplied to the museum; the pages were not opened by the museum, with the founder's approval of 21 November 2026, and each record says so. The United Nations and UNESCO source records already in the graph are reused.
- Two non-Wikipedia sources now support the academy founding years alongside the Wikipedia pointers: Alhawary, "Arabic Applied Linguistics", Cambridge Handbook of Arabic Linguistics (CUP 2021; Damascus 1919, Cairo 1932, Baghdad 1947) and the Jordan General Budget Department chapter 2502 of 2024 (Jordan Academy of Arabic, 1976). Their records state that the museum has not opened them and that a research file named them for these dates only.
Data (arabiyya-browser, script)
- Six early inscriptions of the Arabic script and its ancestors are now records (class object, tagged arabiyya-browser and arabiyya-script), each with prose in Arabic, English, Urdu, French and Spanish: the Raqosh epitaph JSNab 17 (267 CE, Hegra), the Namara inscription (328 CE, Louvre AO 4083), the trilingual Zabad inscription (512 or 511 CE), the Jabal Usays inscription (528/529 CE, alternative 532/533), the Harran inscription in the Lajah (567/568 CE) and the Dome of the Rock inscription of 72 AH. Each states its date and calendar, find-place and present home, language and script, what it says (a paraphrase), and why it matters for the history of Arabic, and keeps the alternative dates and readings the research summary gives.
- Two overview concepts: aqdam-al-nuqush-al-arabiyya (the six in order, with the gaps between them) and marahil-tatawwur-al-khatt-al-arabi (Kufic, dots and vowel marks, Naskh, Thuluth, Ruq'ah), linking the existing script, dotting and calligrapher records without editing them.
- 25 source records for the pages cited, including the Louvre record AO 4083, Lindstedt's chapter in Muhammad and His Followers in Context (Brill, 2024) and a Western Michigan University dissertation. Every record says plainly that the facts come from a research summary supplied to the museum, that the pages were not opened, and that the founder approved this on 21 November 2026; the dissertation's title is about the Qur'anic Jesus, so its fit to the three Syrian inscriptions is stated as unchecked. Wikipedia is cited only for entity facts.
Data
- The Arabic Language browser's books chapter now places 24 works, the twelve dictionaries and twelve grammar books of a research list of the sixty most-cited works on Arabic, in their field. Seven grammar works get new records (al-Muqtadab, al-Usul fi al-Nahw, al-Mufassal fi San'at al-I'rab, al-Kafiya, al-Ashbah wa-l-Naza'ir fi al-Nahw, Jami' al-Durus al-Arabiyya, al-Nahw al-Wafi) and the author of the last, Abbas Hasan, gets a person record; the seventeen works the graph already held (Kitab al-Ayn, Jamharat al-Lugha, Tahdhib al-Lugha, al-Sihah, Maqayis al-Lugha, al-Mukhassas, Asas al-Balagha, Lisan al-Arab, al-Qamus al-Muhit, Taj al-Arus, al-Mu'jam al-Wasit, Mu'jam al-Lugha al-Arabiyya al-Mu'asira, Kitab Sibawayh, the Alfiyya, al-Ajurrumiyya, Mughni al-Labib, Sharh Ibn Aqil) gain the books tags, their sources and a "Place in the canon" section, and the ten of them that had no prose gain five-language pages. Every record is in Arabic, English, Urdu, French and Spanish.
- 45 source records, one for every documenting page and every Shamela library link the list names (24 Wikipedia pages and 21 Shamela links), are added and cited in the works they support. The list was used with the founder's approval of 21 November 2026 and without re-verification; the museum did not open the pages, and each record says so. Each record says what the museum has and has not read of the book itself.
Data (Arabic Language Browser, books chapter: morphology, prosody, rhetoric)
- Thirty-two new book records join the "books" chapter of the Arabic Language Browser: twelve on morphology (al-Tasrif of al-Mazini, al-Munsif and al-Tasrif al-Muluki of Ibn Jinni, al-Shafiya, Tasrif al-Izzi, al-Mumti, Lamiyat al-Af'al, Sharh Shafiyat Ibn al-Hajib, Marah al-Arwah, Shadha al-Arf, al-Tatbiq al-Sarfi, al-Mughni fi Tasrif al-Af'al), twelve on prosody (from al-Akhfash's Kitab al-Qawafi to Abd al-Aziz Atiq's Ilm al-Arud wa-l-Qafiya) and eight on rhetoric (Ibn al-Mu'tazz's Kitab al-Badi', Qudama's Naqd al-Shi'r, al-Sina'atayn, Miftah al-Ulum, al-Mathal al-Sa'ir, Jawahir al-Balagha, al-Balagha al-Wadiha and al-Maydani's al-Balagha al-Arabiyya). Each has label and summary in Arabic, English, Urdu, French and Spanish, prose of more than 200 words in all five languages, the author linked, and the date and the Shamela availability exactly as the research file gives them.
- Four existing rhetoric records (al-Bayan wa-l-Tabyin, Asrar al-Balagha, Dala'il al-I'jaz, al-Idah) gain a "place in the canon" section in all five languages, the documenting pages as sources and, where missing, the author link. Asrar al-Balagha and al-Idah, which were draft cards with no prose, now have their first prose pages.
- One overview concept, "The mother books of the Arabic sciences", lists all sixty works by field in five tables (title, author, date, in Shamela yes/no/partly) in five languages, linking the record of every book.
- Fourteen new person records for authors the museum did not yet have (for example Izz al-Din al-Zanjani, Ibn Usfur, Abduh al-Rajihi, al-Khatib al-Tibrizi, al-Damamini, Ahmad al-Hashimi, Ibrahim Anis, Safa Khulusi, al-Maydani), each carrying only the death date the research file gives.
- Fifty-nine new source records, one for every documenting page and every Shamela link the file lists for these thirty-six works. Shamela links are bibliographic pointers only. Every record states that the facts come from a research summary, that the museum did not open the pages, and that the founder approved using the summary without re-verification on 21 November 2026; the museum has read pages of only four of these books (al-Idah, Miftah al-Ulum, the Badi' of Ibn al-Mu'tazz, Dala'il al-I'jaz), and those pages are named on their records. Where the file says no standalone Shamela text of Dala'il al-I'jaz was found while the museum's own record cites a Shamela copy it has read, both statements are kept on the page.
Added
- Arabic browser, world chapter: twelve new articles on the status of Arabic outside the Arab states, in five languages: Chad, Eritrea, Mali, Niger, Nigeria, Cameroon, Khuzestan (Iran), Hatay and southeastern Turkey, Israel, Cyprus, Malta and Central Asia. Each gives the legal status, the varieties spoken and the official figures exactly as a research summary supplied to the museum reports them (Mali 2022 census counts, Chad literacy shares 10.3 and 2.1 percent, Israel CBS 18 percent of residents aged 20 and over, Cyprus 19,663 and 4,486 with UNESCO severely endangered, Malta 90.4 and 94.7 percent with UNESCO non-endangered, Central Asia 88,244 and 4,184 with UNESCO definitely endangered). Where the summary found no source or figure, the page says so. Israel and Palestine statements are attributed to their holders, with no verdict.
- Arabic browser, origins chapter: four new articles on how Arabic spread after the 7th century: the conquests and garrison towns, the Arabisation of the dīwāns (Iraq 697, Syria 700, Egypt 705-706, Khurasan 741-742), Arabic in al-Andalus and Sicily with Malta and Maltese and the end of Arabic in Spain, and the Arabisation of the Maghrib (Kairouan 670, pre-Hilalian then Hilalian waves).
- 91 new source records for every page the research files list (Wikipedia pages marked as supporting entity facts only). The museum did not open these pages; the founder approved using the summaries on 21 November 2026. Source records are draft tier.
Added
- Arabic Language Browser: three concept records on diglossia and teaching (Diglossia in Arabic; Classical Arabic and Modern Standard Arabic; Teaching Arabic as a foreign language), each a full article in five languages, with tables of Badawi's five levels, the differences in vocabulary, syntax and style, and the proficiency tests (ACTFL, Hamzah, ALPT, CIMA, AL-ARABIYYA-TEST). Written from a research summary supplied by the founder, whose pages the museum did not open.
- Four grammar reference records built around tables on the root k-t-b: the ten verb forms, derived nouns, the three cases and diptotes, and the dual, plural and numbers three to ten. Page numbers in Ryding 2005, Wright 1896 and al-Ghalayini are the summary's and unchecked.
- A comparison record of Modern Standard Arabic and six dialects (how are you, what do you want, two, three), with a specialist bibliography stating that individual phrases were not attributed to those books.
- A "Common phrases" section (greetings, how are you, what do you want, numbers one to five, with Arabic script, transliteration, MSA equivalent and source per row) added to the Egyptian, Levantine, Gulf, Hijazi, Maghrebi (Moroccan Darija) and Yemeni dialect records in five languages; the Yemeni caveat that the examples are not universal across Yemeni varieties is stated.
- 56 source records for every page the two research summaries list, each marked as not opened by the museum and approved by the founder on 21 November 2026.
Data
- Arabic language browser, poetry and philology (tag arabiyya-scholars): three new concept records, each a full article in Arabic, English, Urdu, French and Spanish: the Mu'allaqat (the seven odes and the three poets added in ten-ode recensions, with tribes and approximate dates), pre-Islamic poetry as linguistic evidence (the Masa'il of Nafi ibn al-Azraq to Ibn Abbas, and Sibawayh's verse citations in al-Kitab with Baalbaki's counts, attributed), and the schools of Basra and Kufa on witnesses, transmission and analogy (with Ibn al-Anbari's 121 points of dispute). Why it matters: readers of the browser can see how verse became the evidence on which Arabic grammar was built.
- Arabic proverbs (tag arabiyya-lisan): a concept on the great proverb books (Abu Ubayd, al-Askari, al-Zamakhshari, al-Maydani, with dates and counts) and ten short articles, one per well-known proverb, each with its story as the books tell it, its classical sources, its meaning and its use; each says the stories are legends as reported, not confirmed history. Slugs begin mathal-.
- Arabic literature (tags arabiyya-scholars, arabiyya-lisan): the eight periods of Arabic literature, the beginnings of the Arabic novel (Haykal's Zaynab, 1913), Naguib Mahfouz and the 1988 Nobel Prize, and language reform in the Nahda (al-Shidyaq, the two al-Yazijis, al-Bustani).
- Translation and Arabisation (tag arabiyya-world): the Abbasid translation movement and how Arabic became the language of science, nine Arabic scientific words in European languages (algebra, algorithm, alchemy, alcohol, zenith, nadir, azimuth, syrup, elixir), and Arabisation of the sciences with the language academies; it states both dates given for the ALECSO coordination bureau (1962 in this summary, 1961 in another).
- 86 new source records for every page the research summaries list (encyclopaedias, dictionaries, journals, library pages, commercial and learning sites alike). All facts come from research summaries supplied to the museum; the pages were not opened by the museum, with the founder's approval of 21 November 2026, and each record says so. Wikipedia pages support entity facts only.
Data
- Arabic language browser: 23 new full-article records, each in Arabic, English, Urdu, French and Spanish (Arabic body 450+ words each), written from five research files supplied to the museum (pages not opened, founder approval 21 Nov 2026). Script and print (tag arabiyya-script): history of Arabic printing (1486 to the Beirut Imprimerie Catholique), printing, type and the Unicode Arabic block, the Arabic keyboard (ASMO 663, Arabic 101/102, Mac), the eleven languages written in the Arabic script (table of added letters, kept or replaced), the letters added in Persian and other languages, and Arabizi. Names (arabiyya-scholars): onomastics overview with the Ibn Khallikan example and an open proposal on structured name fields (awaiting the founder, schema unchanged), ism and kunya, nasab and nisba, laqab and khitab, elements of names (Abd, Ibn, Abu, Umm, al-, Al), how family names developed, and the reference works and authority files. Laws (arabiyya-world): Algeria's 1991 Arabisation law, Morocco's 2011 constitution, 2019 framework law 51-17 and organic law 26-16, Qatar's Law 7 of 2019, the UAE draft law (no enacted text found), and the Arab League, ALECSO and UNESCO. Institutions (group): King Salman Global Academy (2020), Arabic Language Academy in Sharjah (2016), the Arabic Language Centre in Dubai (1980, a private teaching institute), ALECSO and its Rabat Coordination Bureau (1961 here, 1962 in another file, both stated), and the Doha Historical Dictionary project (2013). Why it matters: readers of the browser can follow how Arabic was printed, encoded, typed and written by other languages, how its names work, and how states and organisations protect and plan it.
- 75 new source records for every page the research files list (each states: pages not opened, founder approval 21 Nov 2026).
Added
- Ten new records for the Arabic Language Browser in all five languages, each a full article: how the four Sunni schools treat reciting the Qur'an in translation in prayer, the Friday sermon in a language other than Arabic, whether a translation counts as the Qur'an, and the duty to learn Arabic for worship (positions only, attributed to the named works, no ruling); the first Qur'an translations, translations by language with translators and dates, and why a translation is called an interpretation (al-Zarqani, al-Shatibi, Ibn Taymiyya, Pickthall); and, from the classical commentaries, what "clear Arabic tongue" means, the dispute over non-Arabic words, and whether inimitability lies in the text or in sarfa (al-Nazzam). They matter because readers can see each school's or author's view under its holder, with the work it comes from, and no verdict.
- 50 source records: one for every page the research summaries list (48) plus the two research summaries themselves; each says the pages were not opened and page numbers are unchecked.
Changed
- The ten qira'at, the seven ahruf, Ibn Mujahid's selection of the seven, and the conditions of a sound reading gained sections: the ten readers with their narrators, death dates and features, the dialects linked to the seven ahruf with the other views, and the main works; they now also carry the arabiyya-browser and arabiyya-ijaz tags.
Added (structured names on person records)
- A person record can now carry its classical Arabic name as structured data (`name_parts`). The founder approved the proposal on 21 November 2026. Each element of the name (given name, kunya, line of descent, epithet, honorific title, nisba, ancestral designation, the name the person is known by, spellings and source-attributed variants) can hold several values, each with its own sources, a certainty (attested, disputed or uncertain) and, for a nisba, an optional link to a place, tribe, occupation, dynasty or school that does not claim birth there or descent from it. The field is optional and no record is invalid without it. The validator checks that every cited source and linked record exists. Ibn Khaldun is the first record to use it. The shape is described in
docs/data/name-parts.md. The site does not show it yet.
Tests
scripts/name-parts.test.tschecks the schema for the new field: the Ibn Khaldun record passes, several values and a link pass, and an unknown element, an unknown certainty, a malformed link and a missing value fail.
Added (name card)
- A person's page now shows a name card when the record carries structured name parts. The card lists the classical Arabic name element by element (given name, kunya, descent, epithet, honorific title, affiliation, ancestral name), each with its transliteration, gloss, a "disputed" or "uncertain" mark where the sources are not sure, and, for an affiliation linked to a place, tribe, occupation, dynasty or school, a link to that record. A sentence under the card says a link shows an association only and never that the person was born there or descended from it. The card is plain HTML and ships in Arabic, English, Urdu, French and Spanish. Only pages of people whose records have
name_partschange (Ibn Khaldun today), so the deploy stays small.
v2.26.1 — 2026-11-20
Data
- Assabile harvest register: 9 episodes updated across 2 series, 8 transcripts recorded (0 episodes added). The register now shows which Assabile episodes are on the drive and which machine transcripts are finished, so later harvest work starts from the recorded state. No mined episode's entry was changed beyond its transcript and drive lines.
- Assabile harvest register, second sync of 21 November 2026: 8 episodes updated across 2 series, 8 machine transcripts recorded (0 episodes added). Five episodes of the Ahkam al-Janaiz series (Mohamed Hassan) and three of Wa Nalqa al-Ahibba (Amr Khaled) now show their transcripts and drive state in the register. Mined episodes were not changed beyond those lines.
- Assabile harvest register, third sync of 21 November 2026: 7 episodes updated across 3 series, 7 machine transcripts recorded (0 episodes added). The register now shows the transcripts finished since the last sync, and the transcript state of the affected series, so later harvest work starts from the recorded state. Episodes not yet mined had their entries updated only in the transcript and drive lines; no mined episode's entry was changed.
- Assabile harvest register, fourth sync of 21 November 2026: 14 episodes updated across 3 series, 14 machine transcripts recorded (0 episodes added). The register now shows the transcripts finished since the last sync and the transcript state of the affected series (two series now show complete, one partial), so later harvest work starts from the recorded state. Only episode lines and the transcript and count fields of the three series were changed; no mined episode's entry was changed beyond its drive lines.
- Assabile harvest register, fifth sync of 21 November 2026: 6 episodes updated across 1 series, 6 machine transcripts recorded (0 episodes added). Episodes 5 to 8 of the al-Sira al-Nabawiya series by Mohamed Hassan now show their machine transcripts (pass5-qwen) and the state transcribed, ready for mining.
- Assabile harvest register, sixth sync of 21 November 2026: 3 episodes updated or added across 2 series, 3 machine transcripts recorded (1 episode added). Episodes 13 and 14 of the al-Sira al-Nabawiya series by Mohamed Hassan now show their machine transcripts and the state transcribed; episode 8 of Stories from the Quran 2 by Amr Khaled is added to the register as a video on the drive, not yet transcribed.
- Assabile harvest register, seventh sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 15 of the al-Sira al-Nabawiya series by Mohamed Hassan and episode 8 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, eighth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 16 of the al-Sira al-Nabawiya series by Mohamed Hassan and episode 14 of Mojadedoun 1 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, ninth sync of 21 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 17 and 18 of the al-Sira al-Nabawiya series by Mohamed Hassan and episode 13 of Mojadedoun 1 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, eleventh sync of 21 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 1 and 2 of al-Tarbia li-Ma'atha by Mohamed Hassan and episode 9 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, twelfth sync of 21 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 3 and 4 of al-Tarbia li-Ma'atha by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, thirteenth sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 10 of Stories from the Quran 2 by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, fourteenth sync of 21 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 5 and 6 of al-Tarbia li-Ma'atha by Mohamed Hassan and episode 11 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifteenth sync of 21 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 1 and 2 of el-Sabaa el-Mubikat by Mohamed Hassan and episode 12 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixteenth sync of 21 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 4 and 5 of el-Sabaa el-Mubikat by Mohamed Hassan and episode 13 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, seventeenth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 6 of el-Sabaa el-Mubikat by Mohamed Hassan and episode 14 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, eighteenth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 1 of Haqiqat al-Tawheed by Mohamed Hassan and episode 15 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, nineteenth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 2 of Haqiqat al-Tawheed by Mohamed Hassan and episode 16 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, twentieth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 3 of Haqiqat al-Tawheed by Mohamed Hassan and episode 17 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, twenty-first sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 18 of Stories from the Quran 2 by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, twenty-second sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 4 of Haqiqat al-Tawheed by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, twenty-third sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 19 of Stories from the Quran 2 by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, twenty-fourth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 5 of Haqiqat al-Tawheed by Mohamed Hassan and episode 20 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, twenty-fifth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 6 of Haqiqat al-Tawheed by Mohamed Hassan and episode 21 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, twenty-sixth sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 22 of Stories from the Quran 2 by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, twenty-seventh sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 3 of el-Sabaa el-Mubikat and episode 7 of Haqiqat al-Tawheed, both by Mohamed Hassan, now show their machine transcripts and the state transcribed.
- Harvest register: Jami al-Tawarikh (Rashid al-Din) added, both volumes OCR'd with Apple Vision (21 November 2026). Two PDF volumes from an archive.org scan are on the drive; the full text of both is read page by page (412 and 240 pages). Mining has not started yet.
- Assabile harvest register, twenty-eighth sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 23 of Stories from the Quran 2 by Amr Khaled now shows its machine transcript and the state transcribed.
- Harvest register correction: the Jami al-Tawarikh scan holds volume 2 parts 1 and 2, not volume 1. The title pages read المجلد الثانى; volume 1 is not in this archive item. The register entry now says so (21 November 2026).
- Assabile harvest register, twenty-ninth sync of 21 November 2026: 6 episodes updated across 2 series, 6 machine transcripts recorded (0 episodes added). Episodes 8 to 11 of Haqiqat al-Tawheed by Mohamed Hassan and episodes 24 and 25 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, thirtieth sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 26 of Stories from the Quran 2 by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, thirty-first sync of 21 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 12 and 13 of Haqiqat al-Tawheed by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, thirty-second sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 14 of Haqiqat al-Tawheed by Mohamed Hassan and episode 27 of Stories from the Quran 2 by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, thirty-third sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 15 of Haqiqat al-Tawheed by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, thirty-fourth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 16 of Haqiqat al-Tawheed by Mohamed Hassan and episode 1 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, thirty-fifth sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 17 of Haqiqat al-Tawheed by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, thirty-sixth sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 18 of Haqiqat al-Tawheed by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, thirty-seventh sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 2 of Al-Khota al-Habib by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, thirty-eighth sync of 21 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 19 and 20 of Haqiqat al-Tawheed by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, thirty-ninth sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 21 of Haqiqat al-Tawheed by Mohamed Hassan and episode 3 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, fortieth sync of 21 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 22 and 23 of Haqiqat al-Tawheed by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, forty-first sync of 21 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 24 of Haqiqat al-Tawheed by Mohamed Hassan and episode 4 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, forty-second sync of 21 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 1 of Jibril wa al-Nabi by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, forty-third sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 5 of Al-Khota al-Habib by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, forty-fourth sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 10 and 11 of Jibril wa al-Nabi by Mohamed Hassan and episode 6 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, forty-fifth sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 12 and 13 of Jibril wa al-Nabi by Mohamed Hassan and episode 7 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, forty-sixth sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 14 of Jibril wa al-Nabi by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, forty-seventh sync of 22 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 15 of Jibril wa al-Nabi by Mohamed Hassan and episode 8 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, forty-eighth sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 17 of Jibril wa al-Nabi by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, forty-ninth sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 16 and 18 of Jibril wa al-Nabi by Mohamed Hassan and episode 9 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, fiftieth sync of 22 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 19 and 20 of Jibril wa al-Nabi by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifty-first sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 21 and 22 of Jibril wa al-Nabi by Mohamed Hassan and episode 10 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifty-second sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 23 of Jibril wa al-Nabi by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, fifty-third sync of 22 November 2026: 2 episodes updated across 2 series, 2 machine transcripts recorded (0 episodes added). Episode 24 of Jibril wa al-Nabi by Mohamed Hassan and episode 11 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifty-fourth sync of 22 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 25 and 26 of Jibril wa al-Nabi by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifty-fifth sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 27 and 28 of Jibril wa al-Nabi by Mohamed Hassan and episode 12 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifty-sixth sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 29 of Jibril wa al-Nabi by Mohamed Hassan now shows its machine transcript and the state transcribed.
- Assabile harvest register, fifty-seventh sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 13 of Al-Khota al-Habib by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, fifty-eighth sync of 22 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 30 and 31 of Jibril wa al-Nabi by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, fifty-ninth sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 32 and 33 of Jibril wa al-Nabi by Mohamed Hassan and episode 14 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixtieth sync of 22 November 2026: 1 episode updated in 1 series, 1 machine transcript recorded (0 episodes added). Episode 15 of Al-Khota al-Habib by Amr Khaled now shows its machine transcript and the state transcribed.
- Assabile harvest register, sixty-first sync of 22 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 34 and 35 of Jibril wa al-Nabi by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixty-second sync of 22 November 2026: 5 episodes updated across 3 series, 5 machine transcripts recorded (0 episodes added). Episodes 36 and 37 of Jibril wa al-Nabi, episodes 1 and 2 of Khatirat al-Fajr, both by Mohamed Hassan, and episode 16 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixty-third sync of 22 November 2026: 4 episodes updated in 1 series, 4 machine transcripts recorded (0 episodes added). Episodes 3 to 6 of Khatirat al-Fajr by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixty-fourth sync of 22 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 7 and 8 of Khatirat al-Fajr by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixty-fifth sync of 22 November 2026: 3 episodes updated across 2 series, 3 machine transcripts recorded (0 episodes added). Episodes 9 and 10 of Khatirat al-Fajr by Mohamed Hassan and episode 17 of Al-Khota al-Habib by Amr Khaled now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixty-sixth sync of 22 November 2026: 2 episodes updated in 1 series, 2 machine transcripts recorded (0 episodes added). Episodes 11 and 12 of Khatirat al-Fajr by Mohamed Hassan now show their machine transcripts and the state transcribed.
- Assabile harvest register, sixty-seventh sync of 22 November 2026: 3 episodes updated in 1 series, 3 machine transcripts recorded (0 episodes added). Episodes 13 to 15 of Khatirat al-Fajr by Mohamed Hassan now show their machine transcripts and the state transcribed.
Changed
- Rule change for one lane: audio of the YouTube channel @muftimenkofficial may be downloaded for local transcription (founder ruling, 21 November 2026). The ruling is recorded in the sources policy and in the three documents that used to forbid YouTube downloads (plan 25, the audio and video harvest method, plan 15), scoped to that channel alone. The audio and transcripts stay in the harvest workspace and are never published; the museum publishes only its own wording, cited by timestamp, with a link to the video.
- The YouTube lecture harvest can run its claims pass without the judge step, and the judge can follow later. The harvest script gains two options: one produces claims and headers and holds each video at
claims_done; the other runs the judge on videos already at that point. A claim is not counted as checked until the judge has run, so nothing moves to a record on the first pass. No site content changed.
Added
- Muhammad Abduh's Risalat al-Tawhid, the treatise on the science of Tawhid, now has three concept entries in Arabic, English, Urdu, French and Spanish, drawn from the body of the book: good and ugly acts, God's knowledge and human choice (with the author's view labelled as his), and the general mission of the messengers and the miracle. The source and work records now say that the introduction and the body pages at scan pages 63 to 70 and 82 to 84 have been read; the rest of the book is not yet mined.
- How to Savour Prayer (Kayfa tatalaththadhu bi-al-salat), a video lecture series by Mishari al-Kharraz held in the Assabile archive, now has an entry in Arabic, English, Urdu, French and Spanish. It is written from the first three episodes of the machine transcript and covers presence of heart, understanding, hope and awe in prayer. The speaker's own figures and reports are marked as his and as needing a source, and passages the transcript renders unclearly are flagged rather than guessed.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 53 and 54 added in Arabic, English, Urdu, French and Spanish as drafts. Episode 53 covers the wars against the apostates after the Prophet's death, the claimants to prophecy (Tulayha, Sajah, Musaylima), the eleven armies Abu Bakr sent out, Khalid's campaign against Tulayha, the return of Tulayha and the capture of Uyayna ibn Hisn, and the killing of Malik ibn Nuwayra as the speaker tells it. Episode 54 covers the battle of Yamama against Musaylima: the Hanifa army, the death of Habib ibn Zayd, the tribal ranks, Abu Dujana and Zayd ibn al-Khattab, the walled garden, the peace with Mujjaa, and Khalid's later marriage. Every point is the speaker's account, with timestamps; the speaker's figures are marked as unchecked; the conflicts inside the transcript and with the record are stated; garbled and present-day passages are left out.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 49, 51 and 52 added in Arabic, English, Urdu, French and Spanish as drafts. Episode 49 is Tufayl ibn Amr al-Dawsi (his conversion on hearing a passage of the Qur'an, the light between his eyes, the conversion of his people Daws, the Prophet's prayer for them, and his death as a martyr at Yamama). Episodes 51 and 52 continue the Khalid ibn al-Walid series: his conversion on the road to Medina, Mu'tah and the withdrawal, the conquest of Mecca, Hunayn, Tabuk and Dumat al-Jandal, and Banu al-Harith. Every point is the speaker's account, with timestamps; figures and ages are marked as unchecked; the conflicts between the transcript and the record are stated; garbled passages and present-day framings are left out.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 45 and 46 added in Arabic, English, Urdu, French and Spanish as drafts. Episode 45 is Jabir ibn Abdullah (the second pledge of al-Aqaba, the lot for Uhud, his father's debts and death, the flood that uncovered the grave, the army fed from a handful of barley, and the camel sold to the Prophet). Episode 46 is Sumayya, Yasir and Ammar ibn Yasir (the first family to embrace Islam openly, Sumayya as the first martyr, Ammar's torment, Badr, and Umm Umara at Uhud and against Musaylima). Every point is the speaker's account, with timestamps; the figures and ages are marked as unchecked, the contested war account is framed as the speaker's, and the garbled passages are left out.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 36, 37 and 38 added in Arabic, English, Urdu, French and Spanish drafts. Episode 36 is Abu Ayyub al-Ansari (the Pledge of Aqaba, the Prophet's seven months in his house, the manners of visiting and asking permission, and his last campaign at Constantinople). Episode 37 is Abu Hurayra (his name, his years with the Prophet, the People of the Suffa, his memory of hadith as the speaker presents it, his mother's conversion and his worship). Episode 38 is Bilal ibn Rabah (his torture and purchase, the call to prayer, Badr and the conquest of Mecca, his last years in Damascus). Each statement is cited to the recording by timestamp; the speaker's dates, figures and reports are marked as needing a source. Machine transcripts, not listened to.
- The children's podcasts now have their full record in the repository. A new engineering document (65 · Podcast pipeline) sets out how "Hudhud's Story Time" (English) and «وقت حكايات هدهد» (Arabic) are made: the ElevenLabs voices, and why Arabic needed a grandfather voice of its own; scripts drawn only from the museum's own records; the picture-book videos; the pre-publish gate every episode must pass — a transcription checked line by line against the script, a loop scan, language checks down to five-second windows, a glyph preflight, and a page-by-page inspection — and the four defects that gate caught and fixed, among them a line one Arabic episode spoke in English. A companion outreach document (
docs/outreach/podcast-series.md) catalogues all twenty episodes with the museum record each story came from and its YouTube link, records where the series stand (both complete on YouTube on 20 November 2026; the museum's own podcast feed goes live at the next deployment, with Apple Podcasts and Spotify to follow), and fixes the one sentence outreach may use about them. - The sack of Baghdad in 1258 now says what the TDV İslâm Ansiklopedisi says about it, in Arabic, English and Urdu. The event page gains a section with the encyclopaedia's accounts: Hulagu's march from Hamadan on 9 Rabīʿ I 655 (27 March 1257); the figures for the dead (800,000 to 2,300,000) and the differing reports on the caliph's killing and on Ibn al-ʿAlqamī; the books thrown into the Tigris and the river that ran for days the colour of ink, which the article on al-Mustaʿṣim presents as a report found in the sources; the caliph's own library, Khizānat al-Mustaʿṣim; and the date of the caliph's death, 20 Muḥarram 656 (27 January 1258), against the 10 Ṣafar of the programme account already on the page. The House of Wisdom page gains a section on its end: the encyclopaedia says Hulagu burned and destroyed it in 1258 after more than five hundred years. The page also lists its other names (Siege, Sack and Fall of Baghdad, 1258) so that a search finds it. Four encyclopaedia articles are new source records, and the encyclopaedia's own judgment that Islamic civilisation began to decline from this date is stated as its judgment only.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episode 29 added, the third Aisha bint Abi Bakr episode, in Arabic, English, Urdu, French and Spanish drafts — the recap of the outings and the night prayer; the speaker's remarks on the privacy of the Prophet's household and the shared bathing vessel; the long women's-husbands hadith with the Abu Zar' and Umm Zar' story; the knot-in-the-rope exchange and the answer on why love lasts; the household's poverty and the wives' request for more maintenance that led to the verses of the choice, with Aisha's answer; the Prophet's last days (the nights, the nursing in her house), the siwak, the head on her chest, the angel-of-death report and the announcement of the death with Abu Bakr's words; her forty-odd years afterwards on the speaker's count, the first school, the tayammum occasion, the saying of Abu Musa and the nephew's question about medicine, the saying about eloquence, her night weeping over one verse, the Arafa fast, the perfumed coins; the studio participants' comments on marriage and on women's place; her part at Uhud and the Trench on his telling; her giving up the burial place to Umar, her wish to be buried with the other wives, and her death as he gives it with Ibn Abbas's glad tidings and the funeral; and the announcement that the ifk affair is kept for a separate episode and that Hafsa bint Umar comes next; each statement cited to the recording by timestamp, the contested material (the argument for the young bride, the ranking among narrators, the ninety-percent claim, the prayer-interruption question, the participants' opinions, the address to girls, love as mission, what her scholarship shows) kept under perspectives and attributed to the speaker or the participants, the museum's readings of garbled names and every unchecked figure listed in the 'left out' sections, one contemporary political example omitted; machine transcript, not listened to.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 30 and 31 added in Arabic, English, Urdu, French and Spanish drafts. Episode 30 is the slander affair (hadith al-ifk) as the speaker tells it: the return from Bani al-Mustaliq, the lost necklace, the rumour, the consultations, the revelation and the lessons, each statement cited to the recording by timestamp; the Qur'anic verses are linked, not typed. Episode 31 is Hafsa bint Umar: her family and marriages, her widowhood, Umar's search for a husband for her, the Prophet's marriage to her, the honey episode, the withdrawal from the wives and her death. Dates and ages are the speaker's, marked as needing a source; the contested material is kept under perspectives with no verdict. Machine transcripts, not listened to.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 39, 40 and 41 added in Arabic, English, Urdu, French and Spanish as drafts. Episode 39 is Hudhayfa ibn al-Yaman (his choice of the Ansar, Badr and Uhud, his father, the night mission to the enemy camp at the Trench, the names of the hypocrites, and his governorship and death). Episode 40 is Sa'd ibn Mu'adh (the Throne shaking at his death, his conversion through Mus'ab ibn Umayr, Badr and the Trench, the judgement on Banu Qurayza as the speaker gives it, and his death). Episode 41 is Suhayl ibn Amr (his son Abd Allah, Badr, the treaty of Hudaybiyya and its clauses, the Conquest of Mecca, and his conversion and death). Every point is the speaker's account, with timestamps; the figures and the disputed judgement are marked as unchecked.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 42, 43 and 44 added in Arabic, English, Urdu, French and Spanish as drafts. Episode 42 is Abbas ibn Abd al-Muttalib (the uncle who lived to see the Prophet's death, the second pledge of al-Aqaba, Badr and the captives, the Conquest of Mecca and the Abu Sufyan episode, Hunayn, his prayer, and his death). Episode 43 is Ja'far ibn Abi Talib (the first emigration to Abyssinia, his five-point account before the Negus, the recitation of the opening of Surat Maryam, his return, and his martyrdom at Mu'ta). Episode 44 is Asma bint Abi Bakr (the Hijra and the carrying of food to the cave, Abu Jahl's visit, her generosity, her counsel to her mother, and her son Abdullah ibn al-Zubayr). Every point is the speaker's account, with timestamps; the figures, the ages and the disputed readings are marked as unchecked, and the garbled passages are left out.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 57, 58 and 59 added in Arabic, English, Urdu, French and Spanish as drafts. Episode 57 covers Khalid ibn al-Walid's march to Syria, the four armies, the desert crossing and the spring, the battle of Yarmuk and the single command, Umar's removal of Khalid, the entry into Damascus and Khalid's death. Episodes 58 and 59 cover the martyrs of Uhud, in two groups, as the speaker tells them. Every point is the speaker's account, with timestamps; the speaker's figures are marked as unchecked; the conflicts inside the transcripts are stated and not decided; garbled and present-day passages are left out.
- How to Savour Prayer (Kayfa tatalaththadhu bi-al-salat), a video lecture series by Mishari al-Kharraz held in the Assabile archive, now covers episodes 4 to 6 as well, in Arabic, English, Urdu, French and Spanish. Episode 4 is on love of God, His beauty and the sweetness of the love that accompanies fear and hope; episode 5 on God's good treatment of His servants and the shame of forgetting it; episode 6 on the blessing of favours and of inner and outer security, and on shame before God. Every point is the speaker's own account, attributed to him, with timestamps; his figures and his medical and factual claims are marked as needing a source, reports without a named chain are flagged, and passages the transcript renders unclearly are flagged rather than guessed. Passages outside the subject of the series are left out.
- How to Savour Prayer (Kayfa tatalaththadhu bi-al-salat), a video lecture series by Mishari al-Kharraz held in the Assabile archive, now covers episodes 7 to 9 as well, in Arabic, English, Urdu, French and Spanish. Episode 7 is on distraction in prayer, the reward that follows humility, and the speaker's 'magic key' of speaking to God in prayer; episode 8 on the call to prayer and what follows it; episode 9 on five kinds of people in prayer and on the congregation. Every point is the speaker's own account, attributed to him, with timestamps; his figures and his claims are marked as needing a source, reports without a named chain are flagged, and passages the transcript renders unclearly are flagged rather than guessed. Verses he cites without a verse number are not written out, and no passage link is added.
- How to Savour Prayer (Kayfa tatalaththadhu bi-al-salat), a video lecture series by Mishari al-Kharraz held in the Assabile archive, now covers episodes 10 to 12 as well, in Arabic, English, Urdu, French and Spanish. Episode 10 is on ablution: its intention and reward, the doubts that trouble some worshippers, and the speaker's account of how the Prophet performed it. Episode 11 is on the sins that the speaker says ablution removes and on the siwak. Episode 12 is on covering the awrah, adornment for God and preparing for the opening takbir. Every point is the speaker's own account, attributed to him, with timestamps; his statistics and claims are marked as needing a source; reports without a named chain are flagged; passages the transcript renders unclearly are flagged rather than guessed. Legal rulings on points where the schools differ are not written, and are listed for the founder. Verses he cites without a verse number are not written out, and no passage link is added.
Added
- **The last four dictionaries of the research list are in the Dictionaries Browser: the Sharjah Historical Dictionary, the Cairo Academy's *al-Muʿjam al-Kabīr*, *Muʿjam matn al-lugha* and the Dictionary of the Contemporary Arabic Language.** Each is described and linked, not hosted. The Sharjah record is built from the project's own pages, read on 20 November 2026: it spans five eras from pre-Islamic inscriptions to modern use, is arranged by roots, and its statistics page then showed 12,009 roots, 78,259 words and 410,062 historical attestations (the home page of the same site showed smaller figures, which the record notes); its rights are reserved by the Arabic Language Academy in Sharjah. *Matn al-lugha* (Ahmad Rida, 5 volumes, 1958–1959) and Ahmad Mukhtar Umar's dictionary (4 volumes, 2008 edition of 3,367 pages) come from Wikidata with the revisions recorded, and each author has a record with dates. *al-Muʿjam al-Kabīr* rests on the TDV account of the Academy's plan and its 500-page specimen volume of 1956. The Doha and Sharjah records show that the research list's earlier claim of "127 volumes" for Sharjah is not what the project's own site states. The two namesakes called Aḥmad Riḍā are listed so that the validator keeps them apart.
Added
- Lane's Lexicon and the Doha Historical Dictionary are in the Dictionaries Browser, described and linked, not copied. *An Arabic-English Lexicon* has a record with its author, Edward William Lane (17 September 1801 – 10 August 1876, from Wikidata with its revision recorded): eight parts, an 1863 edition by William and Norgate, recorded as public domain, running only to the letter qāf with later roots present but not complete, and where it can be read (Perseus, the Tyndale archive, an Internet Archive text version whose rights are not stated). The Doha Historical Dictionary has a record built from its own terms of use, quoted, which allow reuse only for personal non-commercial purposes, forbid derivative works without written permission and forbid commercial or AI-training use without permission, and from Arabica's note that Ramzi Baalbaki heads its Scientific Council; its own about and statistics pages load by script and could not be read, so the record says no more than that. The browser gains a section for historical dictionaries.
Added
- Ramzi Baalbaki's overview of the classical Arabic dictionaries, from Arabica, is now in the records. Three more Arabica entries were read and registered (the overview, the entry on the word *muʿjam*, and the entry on al-Bakrī). From the overview: al-Jawharī (d. about 400 AH / 1010) was the first to make the alphabetical order of the root letters the only criterion of the rhyme system, probably following his uncle Isḥāq al-Fārābī, with al-Bandanījī (d. 284/897) before them considering only the last letter; *Lisān al-ʿArab*'s five sources and Ibn Manẓūr's reasons for them (collection and arrangement); that *al-Qāmūs* has about three times the proper names of *Lisān* and twenty times those of *al-Ṣiḥāḥ*; that *Tāj al-ʿArūs* took fourteen years to dictate, ended in 1188/1774, and has 11,978 roots against 9,273 and 5,618; and short notes on *Asās*, *Maqāyīs*, *Jamhara*, *al-Mukhaṣṣaṣ* and *Dīwān al-adab*. The earlier *Ṣiḥāḥ* sentence, which stated only that al-Bandanījī came first, is replaced by this fuller account. The browser's home gains a short note on what *muʿjam* and *qāmūs* originally meant: *muʿjam* meant a dotted book and was first used as a book title by the hadith scholars, and *qāmūs* meant "the ocean".
Added
- Ten more sources for the dictionaries, opened and checked: seven Arabica entries, two library catalogue records and one manuscript catalogue page. Arabica, the Doha Institute's Arabic encyclopaedia, supplies scholarly profiles by named specialists; the National Library of Tunisia's catalogue gives the Bulaq printings of *Lisān al-ʿArab* (1883–1889, 20 parts in 10 volumes) and *al-Qāmūs* (1856, 2 volumes); the al-Raqq al-Manshur catalogue lists 302 manuscripts of *al-Ṣiḥāḥ*. New in the records: *Kitāb al-ʿAyn* now states Baalbaki's view that its plan is al-Khalīl's by near consensus while his share in the body is unresolved, and the debate on Indian or Greek influence with the voices of Lowe and Haywood, each attributed; *al-Jamhara* notes that Ibn Durayd's stated aim does not match its contents; *Maqāyīs* notes the disputed order with *Mujmal*; *al-Ṣiḥāḥ* notes the claim that al-Bandaniji's *al-Taqfiya* preceded al-Jawharī in rhyme order. Two further sources (HathiTrust, Iranica) refused an automated reader and were not used.
Added
- Entry starts are marked on the pages of al-Ṣiḥāḥ (1,214 pages so far), and the roots can be browsed. A model read the machine text of each page and marked where each entry begins and its root; a script then checked every mark against the exact text, and nothing in the text was changed. Each marked page now shows one paragraph per entry with its root, and
/fr/maajim/sihah/roots/lists the marked roots by first letter with links to the entries (6,382 entries so far). Of 6,607 marks, 95% matched the text exactly and 88 more matched on a shorter start; 225 (3.4%) were dropped because the start was too short or not found, and 14 pages carry no marks. The marks were checked against the page images by eye on three pages (pages 43 and 60 of volume 1, page 300 of volume 2): every entry on them was marked, with one root misread among them. A root can still be misread; the entry on the page is the evidence. The marked pages and roots are also in/api/v1/maajim-sihah.json.
Added
- The Dictionaries Browser lane has a progress file and a written list of which sources it trusts.
docs/delegation/maajim-progress.mdrecords what is built and what is not, the five tests a source must pass before it carries a claim, and a four-tier sort of the 680 sources Gemini cited: those already opened and used (the TDV encyclopaedia, archive.org metadata), those to open and check before first use (manuscript and library catalogues, open journals), reading libraries used to look and not to cite, and sources never cited (AI summary pages, bookshop listings, blogs, Wikipedia prose). It also notes that Gemini's register says its URLs came from search outputs and that it extracted licence text for none of them, so none counts as opened.docs/agents/source-families.mdpoints to it.
Added
- Seventeen more dictionaries in the Dictionaries Browser, from works 26 to 50 of the research list. New records for the dictionaries of Qur'anic and hadith vocabulary (*ʿUmdat al-Ḥuffāẓ*, *al-Fāʾiq*, Abū ʿUbayd's *Gharīb al-Ḥadīth*), of legal vocabulary (*al-Mughrib*, *Taḥrīr alfāẓ al-Tanbīh*, *al-Zāhir*), of philology (*Fiqh al-lugha*, *al-Furūq al-lughawiyya*, *Dīwān al-adab*, *Durrat al-ghawwāṣ*), of loanwords (*al-Muʿarrab*, *Shifāʾ al-ghalīl*), of proverbs (*Majmaʿ al-amthāl*) and of the modern period (*Muḥīṭ al-muḥīṭ*, Wehr, Dozy, *al-Muʿjam al-wasīṭ*). Each states only what the TDV İslâm Ansiklopedisi article says, with dates, editions and arrangement. Among the facts: *Fiqh al-lugha*'s second part was taken from Ibn Fāris's *al-Ṣāḥibī*, after which al-Thaʿālibī made the first part its own book; *Dīwān al-adab*'s supposed Yemen origin is a confusion; al-Muṭarrizī is stated as Muʿtazilī in creed and Ḥanafī in practice with that source named. Five author records are new (al-Samīn al-Ḥalabī, al-Muṭarrizī, Isḥāq al-Fārābī, Abū Hilāl al-ʿAskarī, Shihāb al-Dīn al-Khafājī); Wehr, Dozy and the Cairo Academy's dictionary are modern and in copyright, so they are described and linked, not hosted. The browser gains five more sections and seventeen more source records for the articles used.
Added
- Eleven more dictionaries in the Dictionaries Browser, each with facts checked against the TDV İslâm Ansiklopedisi. New records for *Tahdhīb al-Lugha*, *Jamharat al-Lugha*, *Maqāyīs al-Lugha*, *Mujmal al-Lugha*, *al-Mukhaṣṣaṣ*, *al-Nihāya*, *al-Taʿrīfāt* and *Kashshāf iṣṭilāḥāt al-funūn*, and the earlier cards for *al-Muḥkam*, *Asās al-Balāgha* and *al-Mufradāt* now carry their authors, dates, sources and arrangement. Each work links to its author and, where the encyclopaedia says so, to the dictionaries it draws on (for example *Kitāb al-ʿAyn* to *al-Jamhara*, *al-Maqāyīs* and *al-Muḥkam*). The browser gains three sections: thematic dictionaries, the vocabulary of the hadith, and technical terms. Twelve more source records name each encyclopaedia article, what it was used for and what it does not say.
- The museum no longer states al-Rāghib al-Iṣfahānī's death date as fact. The encyclopaedia entry says the date is very uncertain (classical sources give none; later ones range from 322 to 690 AH, and 502 AH does not fit other evidence it cites). The earlier catalogue's 502 AH is therefore not used; the record cites this source for the uncertainty.
- The museum's scholarly record, written down in the repository: the description paper — *islam.museum: A Citable Open Knowledge Graph and Digital Museum of Islamic History* — was published on Zenodo as a preprint (CC BY 4.0; cite all versions: 10.5281/zenodo.24273662; v1.0: 10.5281/zenodo.24273663), and the dataset record now links back to it (is described by, is cited by). The README gains a "Scholarly record and citation" section, a new engineering document (53 · Scholarly record) sets out the Zenodo archiving flow, the cite-the-concept-DOI policy and where it is implemented, the ORCID/DataCite link, and which indexes carry the dataset as verified on the day; the Cite page gains a section pointing researchers to the paper, and the whitepaper's deposit package in
docs/outreach/whitepaper-2026/is stamped with its DOIs and marked published. The record was refreshed on 20 November 2026 against main of that day: the dataset's daily Zenodo versions now run to v2026.11.20 (10.5281/zenodo.24396188), and the index notes carry the 20 November re-check of the paper's record (14 views, 18 downloads as displayed that day). - Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episodes 33, 34 and 35 added in Arabic, English, Urdu, French and Spanish drafts. Episode 33 is Zaynab bint Jahsh (her family, her brothers at Uhud, her marriage to Zayd ibn Harithah, the verses that ended adoption) and Juwayriya bint al-Harith and the Banu al-Mustaliq. Episode 34 is Safiyya bint Huyayy (the Jews of Medina, the dream, Khaybar, her marriage) and Umm Habiba's journey to Abyssinia and her marriage by proxy. Episode 35 is Maria the Copt (the letter to al-Muqawqis, Hatib ibn Abi Balta'a's journey, Ibrahim's birth and death) and Maymuna bint al-Harith, the last of the wives. Each statement is cited to the recording by timestamp; the speaker's dates and figures are marked as needing a source. Machine transcripts, not listened to.
Changed
- The UNESCO–Sharjah 2027 outreach papers now record the campaign as it stands on 20 November 2026, refreshed against main of the same day. In
docs/outreach/unesco-2027/, the endorsers list, the state of play, the letters and the eligibility memo now show: the founder sent L1, L2, L3 and the IRCICA thank-you himself on 15 November; Professor Hoyland declined on 16 November and that thread is closed, his earlier sentence standing as a personal remark only; Professor An-Na'im agreed to nominate and his signed page is awaited; the Prize Secretariat clarified the nominator rules on 18 November and signalled that the 30 November deadline is expected to be extended; IFLA, ICOMOS and the Muslim World League are parked, and the cold letters (Arab Thought Foundation, Arab British Centre, Professor Bahia Shehab, Biblioteca Islámica) were not sent. The papers now describe the museum as it stands on main today: Arabic first, with the English and Urdu editions live and the French and Spanish editions completed and merged on 19 November 2026, going live at the founder's next deployment; 81,578 entries across 250,362 record pages and 300 pathways on the live site as read on 19 November 2026; and Library and Archives Canada's completed full test crawl of the site (November 2026). Nothing has been filed; the papers say "seeking a nomination", as before. - The English Hudhud stories are all picture-books now, and the catalogue says so. On 20 November 2026 the nine English episodes that were still cover-art videos (Episodes 2–10) were remade as flip picture-books in the first episode's manner — the same tellings, the same titles and the same descriptions, turned page by page — and took their places in the "Hudhud Bedtime Stories" playlist; the cover-art versions stay public on the channel, outside the playlist. The podcast catalogue (
docs/outreach/podcast-series.md) now links each English episode to its picture-book video.
Fixed
- The IndexNow ping now names every sitemap file, read from the fresh sitemap index. The script built its sitemap list from
/sitemap.xml(or named only that file), and a copy of that file made during a deploy stayed in the edge cache for hours: after the production deploy of 20 November it listed 7 of the 11 sitemaps, so the newest ones, which hold the French and Spanish pages, were left out. Every mode ofscripts/indexnow.mjsnow readssitemap-index.xmlpast the edge cache and submitssitemap.xmlplus each file it lists (12 addresses today instead of 1 after a deploy); if the index cannot be read it namessitemap.xmlonly, as before, and the ping still goes. The production-only guard is unchanged. Also documented indocs/indexnow.md: the ping is a job of "Deploy all parts to production" and not of the standalone shared or language deploys. - The parts of the by-language build and deploy now do what the one build and the one deploy did that they had left out. An audit of 21 November 2026 found four gaps. (1)
check:siteandcheck:weightran in no by-language job; a new jobcheck (whole site)puts the shared parts and the five languages together and runs both (on build 44761868478 they pass: 493,860 pages, 0 errors, 86 routes within budget), and the shared deploy refuses a run whose check failed. (2) The page count in the deploy stamp was copied from the previous live stamp: production said 279,334 pages while the site has 493,860, and the Backstage page shows that number; the same job now counts the pages and the shared deploy stamps the count (a run without it keeps the live count and warns). (3) IndexNow, which the one production deploy sent after publishing, is a job of "Deploy all parts to production" again, skipped when any part failed or the repository variable INDEXNOW is off. (4) The documentation states what is still different: noseedinput, five-day artifacts and no pruning. The page count on production stays wrong until the next full deploy of a build that has the new job. - The R2 publish no longer fails on files too big for a Worker. The UAT deploy of 20 November 2026 (run 44758556248) stopped in its second batch with "Memory limit would be exceeded before EOF": the sync Worker unpacks a whole archive in its 128 MB of memory, and the graph export
api/v1/graph/nodes.ndjsonhas grown to 162 MB with the French and Spanish records. The publisher now uploads any file bigger than an archive's 20 MB cap (14 files in the shared build) straight to the bucket with the same S3 tool that stages the archives, with the Worker's own content types, and sends everything else through the Worker as before. The check against the Worker's allowed folders does not apply to these files, because their destinations come from the publish targets, not from outside input. A unit test pins the split. A build made after this change is needed, because the deploy runs the publisher from the built commit. - The test job of the by-language build no longer runs out of memory on a hosted runner. Three builds of
mainin a row (runs 44736140498, 44738042244 and 44742628818) had theirtestjob shut down by GitHub ("The runner has received a shutdown signal") two to three minutes intonpm test, so no build could be deployed. The cause is the memory of the test files that load the compiled graph (549 MB of JSON, about 2 GB in memory each time; one worker was measured at 7.6 GB), which vitest ran three at a time on a 16 GB runner. When the CI variable is set, vitest now runs one test file at a time (vitest.config.ts); nothing is skipped and no assertion changed. The cost is a longer test job in CI; local runs keep every core. - The by-language build of `main` failed at its first step in the English, Urdu and Arabic jobs: the Ritha page imported `scripts/ritha` one folder level short. The path in
site/src/pages/fr/ritha/[...page].astrohad four../where the file needs five (as the Places page has), so the build stopped with "Could not resolve" before building any page; it now resolves, and an English build gets past that step and generates pages. The browser registry also listedmukatabat,munazarat,nar,nature,nubuwatandplacestwice after the Juha and Ritha merges (the build warned "Duplicate key"); the second copy is removed, andjuhastays. - The shared-parts deploy to UAT stopped at the podcast folder, and a guard now keeps that from happening again. The sync Worker's list of folders it may write to (
ALLOWED_PREFIXES) did not havepodcast/, so the publish was refused with "destination not allowed" after the build had been found and thousands of files had been sent;podcast/is now allowed in both Worker environments and excluded from the Pages-only folder in the shared deploy, as it already was in the single deploy. A new test (scripts/r2-registration.test.ts) reads every place that decides where a data folder may go (the publisher, the site Worker's route list and run-first list, both Worker configs, the three deploy workflows' Pages excludes, the root-file rules) and checks every folder ofsite/publicagainst them, so the next new folder fails in a pull request, not minutes into a deploy. The Workers are deployed from the built commit, so a build made after this change is needed before the deploy is started again. - Two Urdu strings of the Dictionaries Browser failed the Arabic-parentheses rule (the pilot note and the Ṣiḥāḥ lede); they now wrap the Arabic parentheticals as the rule requires, and
- Verse topic labels removed from episodes 1 to 6 of How to Savour Prayer (Mishari al-Kharraz, Assabile archive). In Arabic, English, Urdu, French and Spanish, the fourteen bullets that named or restated what a Qur'anic verse says now use the neutral form already used for episodes 7 to 9: the speaker cites a verse, with no verse number given, and the verse is not written. The episode 3 report about the trust and the episode 4 account of Moses are worded the same way, so no verse wording is echoed. Cites and links are unchanged. The note on verses in the left-out list said "six episodes"; it now counts the nine episodes landed (1 to 9) in all five languages.
- إشارة عمر: يرى أن إشارة عمر على أبي بكر بقتل خالد غايتها أن تكون مسأ…" (name the community first, docs/scope.md)
Data
- Assabile harvest register: 8 episodes updated, 2 series, 8 transcripts (21 November 2026). The register sync records the pass-5 transcripts of Mishary al-Kharraz, "Kayfa tataladado bi-salatika" (episodes 6 to 11, audio series), and of Wa Nalqa al-Ahibba by Amr Khaled (episodes 63 and 64, video series); no episode was added. The series headers are recounted from the episode lines.
- Assabile harvest register: 43 episode lines updated, none added, across 6 series (21 November 2026). The register sync sets the transcript (pass5-qwen) on 10 episodes that had none, and records the drive flags for them; 33 episodes already transcribed and mined are marked deleted from the microSD, with the SHA-256 and restore URL kept, and their transcripts stay in place. Episode lines of mined episodes change only in the drive and deleted flags. The Wa Nalqa al-Ahibba lines are included because their packet (PR #7884) is merged.
- Assabile harvest register: 37 fully mined episodes are marked deleted from the microSD (21 November 2026). Each was transcribed, mined and ingested before its audio or video file was removed to free space. The register keeps the SHA-256 and the restore URL for every episode, so the media can be fetched again, and the transcripts are kept.
- Assabile harvest register: 18 episodes added or updated, 3 series, 14 transcripts (21 November 2026). The register sync adds one new episode line and records the pass-5 transcripts of the series it reads, and episodes 57, 58 and 59 of Wa Nalqa al-Ahibba are now marked mined, matching the packet that landed in PR 7891. Only the transcript, mining and on-drive fields changed; no mined episode lost its record.
Added
- Two works of al-Dhahabi now have their own pages: «سير أعلام النبلاء» (`work/siyar-alam-nubala`) and «تاريخ الإسلام» (`work/tarikh-islam-dhahabi`), each with a five-language summary and prose, and linked to their author. Their source records existed but had no work record, so their passages could not be cited.
Data
- The Assabile harvest register records two more transcripts, on two Wa Nalqa al-Ahibba episodes (Amr Khaled). The register sync of 21 November 2026 sets the transcript (pass5-qwen) on episodes 51 and 53, marks them as transcribed, and raises the series' transcribed count from 49 to 51. One series is touched and no book is newly marked OCR complete in this cycle; episodes already mined keep their fields except the drive flags.
- Harvest register: 1,009 Assabile episodes (Omar Abdelkafy's series and others) are marked off the drive (
on_drive: false, deleted 21 November 2026). Each was fully harvested first (transcribed, mined and ingested), carries its SHA-256 and restore URL, and its transcript is kept; the audio or video file is deleted to free about 75 GB on the microSD. - 80 new passages for the Elegy and Death-Announcement Browser: 40 cited excerpts (at most 40 words, machine-read, not corrected) from each of the two books, on elegies, death announcements and condolence, each labelled in five languages.
Data
- The Assabile lecture register gains 12 new series and 220 episodes, and records 4 transcripts. Nine audio series by Mishary Al-Kharraz and Mohamed Hassan and three video series by Amr Khaled now have their episode lines in the harvest register: the file on the drive, its size and checksum, and its download date. Video episodes take the length of their split audio. The four transcripts are marked transcribed; nothing is mined yet, and no existing episode line changed.
- The harvest register now records 141 more Assabile episodes, 3 more series and 2 more transcripts. The register sync of 21 November 2026 adds the new episode lines and their series headers, fills in downloaded files, durations and transcripts, and sets the on-drive state from the queue. In all, 160 episode lines were added or updated. Of already-mined episodes, only the on-drive and deleted states could change.
- The Assabile harvest register records 39 more episode lines, one more series and one more transcript. The register sync of 21 November 2026 adds 34 new episodes, sets 4 durations and applies 1 transcript; four series are touched, one of them new (Mohamed Hassan's taamolat-fi-surat-al-hujurat, nine episodes on the drive). Lines of episodes already mined were not changed.
- The Assabile harvest register gains 160 more episode lines, 2 more series and 1 more transcript, and records 2 Ali ibn Abi Talib books as OCR complete. The register sync of 21 November 2026 adds 160 episodes to the register (series by Mohamed Hassan and others), registers 2 new series and 1 new transcript, and sets the OCR state of *Abghariyat Ali* and of volume one of al-Sallabi's biography of Ali (AAskZad 1158902) to complete. Their failed-page counts are unchanged in the register and are left for the OCR checks to settle.
- The Assabile harvest register gains 182 more episodes, 5 new series and 1 more transcript, and records one more Ali ibn Abi Talib book as OCR complete. The register sync of 21 November 2026 adds 182 episode lines (five new series, three existing series updated), fills 78 durations and records one transcript (pass5-qwen) on a Wa Nalqa episode that was not yet mined. Book OCR: *Ali ibn Abi Talib, His Personality and Times, Part Two* (AAskZad-1158889, 448 pages) is marked complete; the other 35 books in the 21 November OCR list were already marked complete. Nothing already mined changed except its drive flags.
- The Assabile harvest register gains 117 more episode lines in 11 series, and records one Ali ibn Abi Talib book as OCR complete. The register sync of 21 November 2026 adds 117 episodes: nine new series by Muhammad Salih Al-Munajjid and Amr Khaled (112 episodes), and five more episodes in two series already listed. Volume two of *Ali ibn Abi Talib: His Personality and His Era* (AAskZad-1158889) is marked OCR complete; the other 92 OCR-complete books in the queue were already marked. Two episodes of the Wa Nalqa Al-Aheba series, and its series header, are held back this cycle and will land with the packet train that edits them, so no transcript is recorded in this cycle. Counts: 117 episodes, 11 series (9 new), 0 transcripts, 1 book.
- The Assabile harvest register gains 92 more episode lines, 5 new series and 3 more transcripts, and records one more Encyclopaedia of Islam volume as OCR complete. The register sync of 21 November 2026 adds 92 episodes and updates 21 others (113 lines in all), touching 8 series (five of them new); three episodes now carry their transcript. The OCR state of Encyclopaedia of Islam, volume 5 part 3 is set to complete, with the completion stage left at mining because that book is still being mined.
- The Assabile harvest register gains 129 more episode lines and 4 new series, from the pipeline queue of 21 November 2026. Two series by Muhammad Salih Al-Munajjid (Masāʾil Kitāb al-Ṣalāt, 20 audio episodes, and Qiṣaṣ al-Qurʾān, sharh jadid), and an Amr Khaled video series and a cartoon series, now have their episode lines in the harvest register: the file on the drive, its size, checksum and download date. No transcript was added in this cycle and no book was newly marked as OCR complete (the Ali ibn Abi Talib books reported in the queue were already marked). Lines of episodes already mined were not changed.
- The Assabile harvest register gains 75 episode lines in nine series, six of them new, and records two more transcripts; no OCR book is newly finished in this cycle. The register sync of 21 November 2026 adds the episodes of six new series (two by Mohamed Salih Al-Munajjid, four by Amr Khaled) and of three series already listed, with each file's place on the drive, its size and its checksum. Two transcripts are set on the two episodes of the Wa Nalqa al-Ahibba series that already have their transcript in the pipeline queue. Lines of episodes already mined were not changed.
- The Assabile harvest register gains 18 more episode lines, sets 13 durations and records one more transcript. The register sync of 21 November 2026 touches 32 episode lines in three series (18 new lines, 13 durations, 1 transcript). No book is newly marked as OCR complete in this cycle; the eight books still in mining are left as they are.
- The harvest register now records the Assabile lecture archive as the pipeline queue stood on 21 November 2026: 27 episodes added and 46 episode lines updated in all, 4 series touched (3 existing and one new, *Reviving the Islamic Spirit* by Amr Khaled), 2 more transcripts recorded, and 0 books newly finished by OCR. Mined episodes changed only in their on-drive and deleted marks; no record of the museum changed.
- The Assabile lecture register gains 15 more episodes in four series, from the pipeline queue of 21 November 2026. Mojadedoun (1), Daawa li-Taayush, Rehla li-Saada and Reviving the Islamic Spirit each gain their newly downloaded episode lines. No transcript or book OCR state changed in this cycle, and the Wa Nalqa al-Ahibba series was held back from this cycle's register sync, so its lines follow in a later one.
- The Assabile harvest register records 11 more episode lines in three series, and one more transcript; no book is newly finished in this cycle. The register sync of 21 November 2026 fills 10 episode durations (Mojadedoun (1), episodes 5 to 12, and Reviving the Islamic Spirit, episodes 9 and 10) and records Wa Nalqa al-Ahibba episode 48 as transcribed by the pipeline. Episodes 33 to 35 of that series were already in the register and are unchanged. The 124 books in the OCR queue were checked against the register: 109 were already marked complete, and 15 (the Encyclopaedia of Islam volumes and one further book) are being mined, so the register leaves them alone. Data: Assabile harvest register: 0 episodes added, 11 episode lines updated, 3 series, 1 transcript, 0 books OCR done.
- The Assabile lecture register gains three more transcripts and one updated series: Wa Nalqa al-Ahibba by Amr Khaled, episodes 49, 50 and 52 are now marked transcribed by the pass-5 speech model, from the pipeline queue of 21 November 2026. Three episodes updated, one series, three transcripts, and no new episode lines or downloads. The OCR state of the wave-7 books is unchanged: all 125 finished books were already marked complete on main, and the three books that failed in the run (Fihrist of Ibn al-Nadim, Tarikh al-Khulafa and Mirat al-Zaman) stay open.
- Assabile harvest register: 4 episodes updated, 1 series, 4 transcripts. The register sync of 21 November 2026 records the pass5-qwen transcripts of Wa Nalqa al-Ahibba episodes 54 to 57 (Amr Khaled), and each of the four is now marked transcribed. No episode was added.
Data
- Data: Assabile harvest register: Wa Nalqa series reconciled, 37 episodes marked mined (episodes 9 to 31 and 33 to 46, from the landed packets), header recounted from the episode lines (48 transcribed, down from a drifted 55). Episodes 32, 47 and 48 stay flagged, not mined.
Data
- Assabile harvest register (lecture archive): 12 transcript lines added across two series, the Mishary al-Kharraz series "Kayfa tataladado bi salatika" (episodes 12 to 20) and the Amr Khaled series "Wa Nalqa al-Ahibba" (episodes 65 to 67); the register records episode 65 to 67 as transcribed. Wa Nalqa al-Ahibba episodes 49, 51, 52, 53 and 54 are marked mined now, their packets having landed, and the series counts are recounted from its episode lines (58 transcribed, 42 mined). Episodes 47, 48, 50 and 55 keep their state.
- Data: Assabile harvest register: 11 episodes updated, 2 series, 11 transcripts. The register sync of 21 November 2026 records the pass5-qwen transcripts of Mishary al-Kharraz, "Kayfa tataladado bi salatika", episodes 21 to 29, and Amr Khaled, Wa Nalqa al-Ahibba, episodes 68 and 69. No episode is added; each change is a transcript and mining state, and no mined episode changes beyond that.
Added
- The Hudhud's Story Time podcast feeds now carry seasons, and the English feed carries all twenty episodes. The English feed (
/podcast/feed.xml) gains the ten Season 2 episodes, each item tagged with its iTunes season and episode number alongside the ten Season 1 episodes, so Apple Podcasts and Spotify group the series into Season 1 and Season 2 instead of an unknown season. A new Arabic feed (/podcast/feed-ar.xml) carries the ten Arabic episodes as Season 1 with the same tags, ready for its own Apple Podcasts and Spotify shows once it is live. The series catalogue records that the English show was published on Apple Podcasts and submitted to Spotify on 21 November 2026. - Buy Me a Coffee joins the museum's support options. The museum opened its page at buymeacoffee.com/islammuseum on 21 November 2026 (payouts connected through the founder's Stripe account, page currency Canadian dollars). The Support page gains a Buy Me a Coffee card with the service's own icon and a line in each of the five languages, and the landing page's "Donate and support us" list gains the same link, alongside Patreon, PayPal, LaunchGood, GitHub Sponsors, Raisely and GoFundMe. Support stays optional and buys no control, as with every other channel.
v2.26.0 — 2026-11-20
Added
- **The text of *al-Ṣiḥāḥ* is now in the museum, page by page.** The 1865 Būlāq printing, read with Apple Vision, is published as 1,228 records, one for each scan page of the two volumes (646 and 582; covers and library plates are left out). Each record cites its scan, shows the page text and says plainly that it is a machine reading: in a sample of six lines checked by eye against the page image, about one word in five had a wrong or split letter.
/fr/maajim/sihah/lists the pages by volume. The text is not yet divided into entries and nothing beyond that sample has been checked. This follows the founder's ruling of 20 November 2026 (route A, then widened to all the pages), now written intodocs/sources-policy.mdand the decision inbox. - A signature view for the Dictionaries Browser: three doors to one root. The browser's home draws the root ج ر ع filed three ways: by place of articulation in *Kitāb al-ʿAyn*, by last then first radical in *al-Ṣiḥāḥ*, and by first radical in *Asās al-Balāgha*; the Ṣiḥāḥ door links to the scan page where the root actually stands. The browser also has its own thumbnail on the main page, and stays in the Explore menu under Texts.
- Page records of a dictionary show their text. A
passageof unitothernow shows its text and its reliability note on its page; no other page changes.
Data
- **Both volumes of the 1865 Bulaq *al-Ṣiḥāḥ* are now read in full (654 and 590 pages).** The Apple Vision reading of the second volume finished and the register records both outputs with their checksums. The text is still machine output, unchecked, and no entry from it is published yet.
Added
- A new browser for the most famous words of elegy and of announcing a death (`/fr/ritha/`, «متصفح أشهر ما قيل في الرثاء والنعي»). It gathers 607 records from the Qur'an, the Sunnah, the seerah, the reports of the Companions and the Successors and the harvested books: the verses of patience, the Prophet's words «to God belongs what He took and what He gave» and «patience is at the first blow», his announcement of Ja'far, Zayd, Ibn Rawaha and the Negus, the condolences, Hassan's and Ka'b ibn Malik's elegies, and the limits of weeping and wailing. It opens with four mourning banners, a death-notice frame and the famous formulas; it is in the Explore menu, on the landing page, in the API (
/api/v1/ritha.json),llms.txtand the OpenAPI file, in all five languages. - A new record, `concept/ritha-wal-nay`, with prose in Arabic, English, Urdu, French and Spanish, and nineteen existing records on elegy, announcement, condolence and wailing filed into the browser.
Data
- 395 passages harvested for the browser: 74 reports from the seven hadith-api Arabic editions and 321 cited excerpts (at most 40 words, machine-read, not corrected) from 38 OCR'd books, each with its label and summary in five languages.
Tests
- The feature API test covers the new export (
scripts/features-api.test.ts): kinds, corpus sections, a book harvested later, and the three false-positive families (inheritance, source attribution, «نعت» the description).
v2.25.0 — 2026-11-20
Added
- The Juha and Islamic Folklore Browser (متصفح جحا والفلكلور الإسلامي عبر العصور والكوميديا والنكت والطرائف والنوادر), at `/fr/juha/` in all five languages. A new front door for humour in the Islamic tradition, requested by the founder on 20 November 2026: Juha and Nasreddin Hoca, the named wits and fools, the anecdote genre (nawadir), jesting, laughter and smiling, and the ethics of the joke — in the Qur'an, the Sunnah, the seerah, the reports of the Companions and Successors, and the harvested books. The page is a storyteller's gathering unlike any other browser: Juha riding his donkey backwards, the gathering of the wits as medallion seats, a strip of Juha's anecdotes as speech bubbles, a "scale of laughter" showing which kinds of humour each section carries, and the shelf of the anecdote books; below them, six sections with kind chips and a search box. It has a card on the homepage, an entry in the Explore menu, and its own data file at
/api/v1/juha.json(documented indocs/data/README.md, the OpenAPI file andllms.txt). Passages are found at build time by a guarded detector (scripts/juha.ts), so every book harvested later brings its jokes here with no further work. The founder's brief is kept word for word indocs/engineering/64-juha-browser-requirements.mdwith twelve acceptance tests for production; the plan isdocs/plan/72-juha-browser.md. - A new browser: the Rulings Browser — «متصفح الفروض والواجبات والطاعات والمستحبات والمكروهات والمحرمات وصغائر الذنوب» (`/fr/ahkam/`, plan 72, requested by the founder on 20 November 2026). One front door for the degrees of obligation: the obligations, the obligation on every person (fard 'ayn) and on the community (fard kifaya), the acts of obedience, the obligatory, the recommended, the forbidden, the disliked, the minor sins, and what remains of the scale, the permitted and the concession. Its opening view is its own, "the scale of the five rulings": five degrees side by side from the act that must be done to the act that must be left, as al-Ghazali divides them in al-Mustasfa, each with the themes that stand on it. Ten theme pages open with a definition card from the books of usul. The build holds 4,607 records (
npx tsx scripts/ahkam-audit.ts): 130 Qur'an ayahs, 2,500 Sunnah passages, 208 seerah excerpts, 1,735 excerpts from other books and 34 curated records; by theme (a record can carry more than one) 913 obligations, 180 individual obligation, 81 communal obligation, 237 obedience, 929 obligatory, 791 recommended, 1,774 forbidden, 478 disliked, 85 minor sins, 955 permitted and concession. Qur'an verses come from a reviewed table; other passages are found at build time by a guarded word list, so a book harvested later appears by itself; a bare prohibition is not classed as forbidden or disliked, because that is a jurist's reading. Six new records carry prose in Arabic, English, Urdu, French and Spanish (concept/al-ahkam-al-taklifiyya,fard,fard-ayn,fard-kifaya,taa,haram-hukm), with al-Shafi'i's account of the communal obligation cited to al-Risala. The browser has a card on the landing page, an entry in the Explore menu, the dataset/api/v1/ahkam.jsonand its acceptance checklist indocs/engineering/64-ahkam-browser-requirements.md. The museum describes what the texts and the jurists say and issues no ruling. Limits said plainly: the translations are machine-made and the excerpts from scanned books machine-read, none yet read by a person; the full build was not run before the merge, on the founder's instruction; the Explore menu entry changes every page at the next deploy (rewrites-all-pages).
Data
- Juha's anecdotes, one record each. Every anecdote of *Nawadir Juha al-Kubra* (ed. Darwish al-Juwaydi) that passed the text rules is now its own passage record (
passage/nawadir-juha-<n>): the printed Arabic title, a retelling in the museum's own words in Arabic, English, Urdu, French and Spanish, the page and number, and an excerpt of at most forty words. The retellings come from a machine-read scan that was not checked against the page images, and the translations are machine translations no human has read; each record says so. - Humour in the hadith collections and the books. 150 reports of jesting, laughter, smiling and the limits of the joke from the seven hadith collections, and 644 cited excerpts from 24 scanned books — al-Nuwayri's *Nihayat al-Arab* (volume 4), Ibn Qutayba's *'Uyun al-Akhbar*, *al-Mustatraf*, al-Jahiz's *al-Bayan wa-l-Tabyin* and *al-Hayawan*, the *Ihya'*, and the seerah, Companion, biography and history books. Excerpts are chosen by vocabulary, so a few carry the word without being funny; the label says what was detected.
- New and enriched records. Four concepts — jesting (
concept/al-muzah), laughter and smiling (concept/al-dahik), the anecdote (concept/al-nawadir) and the wits and the fools (concept/al-zurafa-wa-al-mughaffalun) — and first full prose in five languages for Juha, Nasreddin Hoca and *Nawadir Juha al-Kubra*, written from the edition's introduction with the editor's claims attributed to him and his contradictory dates for Nasreddin Hoca both given. A source record for the Cairo 1996 scan of *'Uyun al-Akhbar*. Twenty-three existing records (persons, works, places, a concept) carry the browser's tag. The Urdu title of *Nawadir Juha al-Kubra*, which a machine translation had misspelt, is corrected. - 886 hadith reports and 1,368 book excerpts for the Rulings Browser (lane
ahkam): reports from the seven Arabic hadith-api editions and excerpts of at most forty words from 81 books already OCR'd on the harvest drive, led by the books of usul al-fiqh (al-Risala, al-Burhan, al-Mustasfa, al-Muwafaqat, Irshad al-Fuhul), each with labels and summaries in the five languages and amentionsedge to its theme's record. No book was downloaded and no PDF, OCR text or page image entered git.
Tests
- The browser's data file is tested (
scripts/features-api.test.ts): detection of each kind, the section of a Juha collection, a book that does not exist yet landing in "other books", and four false positives that must stay out («طرفة عين», «أرجحا», «الضحاك بن مزاحم», «ابن مغفل»).
v2.24.0 — 2026-11-18
Added
- The by-language trial now covers the whole site, with one start for all of it. The five language deploys published pages and search indexes only. Everything that belongs to no language now has its own build job and its own deploy: the API files, the shared data (media, maps, education files), the feed,
llms.txt, the favicons, the scripts and styles, the site Worker and the Pages project. The sitemap is made by a small job that joins the five languages' own sitemaps with the same code the one build uses; in a sample build the result was the same file, byte for byte. Two new workflows, "Deploy all parts to UAT" and "Deploy all parts to production", take one build and run the shared deploy and then the five language deploys one after another. The one build and the one deploy are unchanged, and nothing of this has run on GitHub yet (docs/engineering/06-deployment.mdsection 9a). - A trial: the site can be built and published one language at a time. The one build writes all five languages on a single GitHub machine, which is more than that machine's documented disk holds, and the build of 18 November on the Spanish release branch stopped that way. A new workflow, started by hand, builds each language on its own machine, five at once, and keeps each language's pages and search index as its own package. Ten new deploy workflows, five for production and five for UAT, publish one language's pages and search index and leave everything else on the site as it is: the other languages, the API, the sitemap, the shared scripts and styles. The existing build and the two existing deploys are unchanged and are still the only way to publish the whole site. Limits a reader should know: a language published alone can be out of step with the others (its language switch may lead to a page another language does not have yet), and the two whole-site checks do not run on a language build. Checked on the founder's Mac with a sample build (about 10,400 pages per language, not the full site): every Spanish page and the Spanish search index of a Spanish-only build are byte-for-byte what the full build writes. Not yet run on GitHub, and nothing has been deployed with it (
docs/engineering/06-deployment.mdsection 9). - Spanish (`es`) as the fifth language of the museum — infrastructure and interface (branch `release/spanish`, proposal `docs/delegation/spanish-language-proposal.md`). The site now builds
/es/beside/ar/,/en/,/ur/and/fr/: the language switcher cycles ar → en → ur → fr → es, each page carries a Spanish hreflang alternate, apagefind-essearch index is built, and the site Worker, the R2 publish list and the deploy workflow know the new folders.i18n/es.yamlcarries the interface in Spanish, translated fromi18n/en.yamlby AI agents and not yet read by a human; dates, eras and month names have Spanish forms. What a visitor does not get yet: no record has Spanish prose, labels or summaries, and the education, kids and tour datasets have no Spanish fields, so those show English (then Arabic) inside the Spanish frame; passage pages have no Spanish translation of the Qur'an (no source is chosen). Spanish stays outside the parity requirements (PUBLISHED_LANGS) until its corpus lands, and no language count indocs/status.md, the site copy or the outreach files is changed. Thespanishlane (docs/agents/lanes/spanish.md, queuescripts/es-queue.ts, progressdocs/delegation/spanish-progress.md) carries the corpus work, in parallel with the French lane. The language rule inAGENTS.mdanddocs/language.mdnow names Spanish for new content. Nothing is live:release/spanishis not merged tomainand nothing is deployed.
Added
- The museum opens in the visitor's own language. Someone who arrives at the bare address
islam.museum/is now taken to the museum in their browser's default language when that language is English, Urdu, French or Spanish; Arabic speakers, speakers of any other language, and browsers without JavaScript get the Arabic page the root has always served. Only the root does this:/ar/,/en/and every deeper address stay where they are, so choosing a language in the header is never undone. A language chosen in the header is remembered: the next visit to the root opens in that language, whatever the browser's own language, until the visitor chooses another. The only thing kept is that language code, in the visitor's own browser storage, never sent anywhere (the privacy page lists it); search-engine crawlers are left on the Arabic root. The change lives in one shared script file (site/public/shell/shell.js), so no page is rewritten. - A research prompt for the planned Dictionaries Browser (`docs/delegation/gemini/dictionaries-browser-prompt-b-history.md`). It asks Gemini for the history, manuscripts, printings, reception and controversies of 40 classical Arabic dictionaries and of their authors, in seven batches, with every claim sourced and unverifiable items marked. It also records what two earlier rounds of research found about where the texts can lawfully come from. Nothing is built or published; the answers will be checked before any record is written.
- A written proposal for the Dictionaries Browser (`docs/plan/72-dictionaries-browser.md`). It sets out what the browser holds (the dictionaries, their authors, their sources, and entries readable in full), the first run of 40 works, what two rounds of research established about where texts can lawfully come from, and the one decision still open: where the readable text comes from. It also records that the museum's own scans of *Lisān al-ʿArab*, *al-Qāmūs* and *al-Ṣiḥāḥ* are too noisy to display. Nothing is built.
- The Dictionaries Browser starts: «متصفح المعاجم» at `/fr/maajim/`. A new page in all five languages lists the museum's records of the great dictionaries in sections (general dictionaries of Arabic, the vocabulary of the Qur'an, geographic dictionaries), with each work's author, and explains where the text of a dictionary will and will not be shown. Nine dictionary records are in it in this first batch, among them *Lisān al-ʿArab*, *Kitāb al-ʿAyn*, *al-Ṣiḥāḥ*, *al-Qāmūs al-Muḥīṭ* and *Tāj al-ʿArūs*; the five general ones now carry dates, named sources and their arrangement, each taken from the Turkish TDV İslâm Ansiklopedisi and cited on the record (for example that Ibn Manẓūr finished *Lisān al-ʿArab* on 22 Dhū l-Ḥijja 689 AH, 26 December 1290, from five named earlier dictionaries). The whole list is also available as
/api/v1/maajim.json, is documented indocs/site/browsers.md,site/public/openapi/v1.jsonandllms.txt, and is found by search. - Eight encyclopaedia source records and two scan records for the dictionaries. Each TDV article that a record relies on is its own source record, with what it was used for and what it does not say (two of them are short cross-references and are used only for one fact each). The two scan records describe the Bulaq 1282 AH / 1865 printing of *al-Ṣiḥāḥ* held by the Internet Archive, which the archive itself marks as not in copyright, with the file sizes and checksums of the copies the museum downloaded.
- A pilot for reading a dictionary in full. The two volumes of the 1865 Bulaq *al-Ṣiḥāḥ* were downloaded and volume 1 is being read with Apple Vision; on a test page the result is far cleaner than the museum's older scans of the same book but still contains errors, so no entry is published yet. The browser page says so, and each entry will show its page and how accurately it was read. The founder chose this route (a pilot from a pre-1925 print, with other dictionaries linked, not copied) on 20 November 2026.
Tests
- The test suite runs again after the Spanish and French pages arrived (`scripts/build-graph.test.ts`). One test file validated the whole repository and also compiled the whole graph twice. With 12,700 more pages the compiled graph alone holds about 3 GB, and the test process ran out of memory, on the build machine and on a laptop alike. The graph tests now have their own file and compile the graph once; the same assertions run, none was dropped and no memory limit was raised.
Added
- The Qur'an can be read with a French and a Spanish translation. A verse page and the sura reader showed a translation of the verse in English (Pickthall) and Urdu (Jalandhari) only. French readers now see Rachid Maach's translation and Spanish readers Muhammad Isa García's, both as published by QuranEnc.com (versions 1.0.3 and 1.0.2), with their footnotes, the translator's name, the version and a link to the source. No French or Spanish translation accepted among Sunni institutions is in the public domain, so these two are shown under QuranEnc.com's own terms: unchanged, credited, with the version stated. They are third-party content, not CC0: they are kept in their own files, outside the museum's data, the frozen Qur'an records and the open API (
docs/data/quran-translations.md). The Arabic text is untouched. Founder's decision, 20 November 2026.
Fixed
- The courses are now built in French and Spanish too, and their language links work. The course and lesson pages were written for Arabic, English and Urdu only, while every one of them offered a French and a Spanish version and the French and Spanish course index listed all 65 courses: 2,686 links to pages that did not exist, which the link check of the full build caught. The lesson text and the interface strings for both languages were already complete, so the pages are now built: 426 more in each language. Course and lesson titles still appear in English there, because the records carry no French or Spanish labels yet.
- Fifty French pages no longer fail the validator. They carried the headings «Pour les Petits» or «Pour les Chercheurs», which the site reads as reading-level sections, while their Arabic, English and Urdu pages have an ordinary closing section. The headings are now «Pour les jeunes visiteurs», «Pour les plus jeunes visiteurs» and «Notes pour les chercheurs», as in the rest of the French pages; the text under them is unchanged.
npm run validateis back to 0 errors. - A link to the hadith narrator Ibn Hamas is whole again in Spanish (
passage/muwatta-31-2597): the machine translation had put only «Ibn» inside the link, which also tripped the content-safety check on the bare second word. - A Spanish viewpoint block may open with any Sunni label that contains «suní» or «suníes» (
scripts/validate.ts), not only one that starts with it, and never one that says «no suní». - Episode 4 of "Hudhud's Story Time" in the podcast feed now points to a repaired recording. The first published audio of "The Woman Who Built a University" was faulty: partway through, the narration fell into a repeating loop and never delivered the middle of the story. The episode has been re-recorded with the same script and the same two voices, checked in full against the script, and the feed entry now carries the new recording (2:41 instead of the faulty 6:18). Anyone who downloaded the old file has the loop; the feed itself now serves the repaired episode.
Data
- The 164 firewall-exception records are translated (founder's ruling, 20 November 2026). These are the records that mention an armed or extremist organisation in a neutral, factual way and that the founder reviewed one by one and kept. They had been left out of the Spanish and French translation until he ruled; he ruled that they are translated like the rest. Their prose pages are now in Spanish and French and their titles and summaries in Urdu, French and Spanish, by the same route as the other records. Each translation was compared with its English source: none names a blocked organisation the English does not. The ruling is recorded in
docs/delegation/extremism-firewall-exceptions.md. Machine translation, not read by a human.
Rolled back
- Translations taken back out of the 164 firewall-exception records. The Spanish and French titles and summaries merged earlier on 20 November (and the Urdu, French and Spanish steps of one exhibit) had also been written into the records listed in
docs/delegation/extremism-firewall-exceptions.md. Those records were reviewed one by one by the founder and are no precedent for new content, and their prose has been held from translation for his ruling; the same now holds for their titles and summaries. The 164 records are back exactly as they were before those merges, and the tool skips them.
Data
- Record titles and summaries in Urdu: 70,339 more texts in about 43,400 records. Two records in five had an Urdu title and summary; now nearly all do. Translated from English by Google Translate's document translation (two files uploaded by the founder), checked by script; blessing formulas the machine added were removed; 463 names it left in Latin letters are still missing. The frozen Qur'an passage records and the held records are untouched. Machine translation, not read by a human.
- Spanish and French catch-up: 245 Spanish and 221 French titles and summaries in records written with quoted keys, which the tool could not fill before.
- Record titles and summaries in French: 112,674 texts in 63,881 records. About one record in eight had a French title and summary; the rest showed English or Arabic to a French reader. They were translated from English by Google Translate's document translation (four files uploaded by the founder) and checked by script, with the same choices as for Spanish: person names keep their English transliteration, titles of sources are left as published, the frozen Qur'an passage records are untouched. Machine translation, not read by a human.
- Record titles and summaries in Spanish: 133,924 texts in 74,497 records. Cards, lists, headings and search showed English or Arabic to a Spanish reader because no record had a Spanish title or summary. They were translated from English by Google Translate's document translation (five files uploaded by the founder) and checked by script; person names keep their English transliteration; titles of sources are left as published; the frozen Qur'an passage records are untouched. About 750 texts the checks refused are still missing. Machine translation, not read by a human.
- The validator and the tests are given 8 GB of memory (`.npmrc`). They read every page and record into memory, about 7 GB now, and the validator stopped with "heap out of memory" at Node's default of about 4 GB once the Spanish texts were added.
- Every exhibit and trail now has its title, summary and steps in Urdu, French and Spanish (318 records). A trail such as «Islam in Thirty Minutes» showed its title in Arabic and its twelve steps in English to an Urdu, French or Spanish reader, because that text lives in the record and had Arabic and English only (19 exhibits also had Urdu). 11,524 short texts were added: translated from English by Google Translate's document translation (the founder uploaded the file by hand), checked by script, and 19 written by Claude where the machine left a name in Latin letters or the English field itself held Arabic. Blessing formulas the machine added after «the Prophet» in Urdu were taken out, as the museum types none. Machine translation, not read by a human. Tool:
scripts/local-mt/data_strings.py. - The museum's pages on single Qur'an verses are now in Spanish and French: 6,350 pages in each language. They hold the museum's own prose about a verse (its place in the surah and, for about 1,700 verses, a summary of al-Tabari's commentary); the verse text itself stays in Arabic, from the frozen Tanzil file. 12,687 pages were translated from English by Google Translate's document translation, the 13 that failed the automatic checks by a Claude agent. Machine translation, checked by script, not read by a human.
- The last very long French records are translated: ten by Google Translate's document translation and four in consecutive parts by Claude agents, one of them begun by Gemini; each file names its translators.
Changed
- New content must come in all five languages, whatever its kind (founder rule, 20 November 2026; `AGENTS.md`, `docs/language.md`). The standing rule asked for Arabic, English, Urdu, French and Spanish on prose and interface strings. It now also names the titles, summaries and every other language block inside data records, the site datasets, the vocabulary files and text written in pages: a new record, field or dataset that lacks one of the five is not finished. Older content stays valid and is being filled.
- Every pull request now goes to `main`; the Spanish release branch is retired (founder's instruction, 19 November 2026). The Spanish and French translation work was collected on
release/spanish. It is now merged intomain, and the lane document, the translation how-to and the shipping scripts of the lane (scripts/local-mt/ship_gt.sh,ship.sh) raise againstmainonly. - How bulk translation is done is written down (`docs/engineering/63-translation-by-document-files.md`, pointer in `AGENTS.md` and the Spanish lane document). One English Word file per round goes through Google Translate's document translation, once for Spanish and once for French; link labels travel as Word hyperlinks and citations as 10-digit numbers, which kept 96–100% of records intact where text markers kept 11–63%. The page records the limits found (about 1.2 million words per file; a usage limit after about fifteen files) and that Arabic as the source was tested and not adopted.
- Translation moves from the local model to Google Translate documents (founder's instruction, 19 November 2026). A pilot showed that a plain-text Word file of about one megabyte (1,644 concept records, about 494,000 English words) is translated by translate.google.com in under a minute, where the local model needs more than a day.
scripts/local-mt/import_gt.pysplits the downloaded file back into records, restores citations and link ids, and applies the same checks as the local translator; 1,588 Spanish and 1,282 French records from the pilot passed and are in, each naming Google Translate in itstranslatorsandprovenance. Records whose citation or link markers Google altered (about 3% in Spanish, 22% in French) were not imported. The local model threads are stopped. Nobody has read these translations. - The site build now clears unused tools off the build machine before it starts. GitHub's build machine is documented with 14 GB of disk, and a build of five languages writes more than that; the build of 18 November on the Spanish release branch most likely stopped for want of disk, though that run printed no reading that proves it. The machine ships with tools this build never uses (the Android SDK, .NET, Haskell, CodeQL, Swift, Docker images). A new first step removes them and prints the free disk before and after. How much it frees is not yet measured: the next build's log will show it. The site itself is unchanged.
- The site build now reports the build machine's free disk and memory while it runs. The build of 18 November on the Spanish release branch stopped after two hours because GitHub lost contact with the machine. Its log simply stops while pages are being written, about 462,000 pages in, with no error message, so nobody could tell whether memory or disk had run out. (An earlier wording of this note said the run left no log; that was wrong, and is corrected here.) The build step now prints both before it starts and every five minutes, and raises a warning on the run page when under 4 GB of disk or 1 GB of memory is left. It only reads; the build itself is unchanged.
- Translation now runs on a local model on the founder's Mac (from wave 11). Spanish and French prose are translated from English by TranslateGemma 27B, served locally with Ollama, under
scripts/local-mt/(a translator that protects citations, link ids and markers and checks every record against its English source, and a supervisor that restarts it). Claude manages the queue, the checks and the merges and no longer writes the translations. The local drafts are freer than the earlier ones and are labelled with the model's name on every record; nobody has read them.
Data
- First Spanish prose: 34,598 records (`content/es/`, branch `release/spanish`). Every eligible record of the canon's top two tiers and 792 object and building records, outside the sacred classes — people such as Umm Salama, al-Nawawi, al-Tabari, Rumi and Mansa Musa, places such as Granada, Qayrawan, Bukhara and Delhi, buildings such as the Dome of the Rock and al-Azhar, events, works, concepts and periods — now have their full prose in Spanish, translated from the English by AI agents and checked mechanically against it (same links, citations, headings and paragraphs; no typed honorific). No human has read them. About 33,400 prose files are still English under
/es/, and every label and summary, and the education, kids and tour datasets are still not in Spanish.
Fixed
- French and Spanish are now named wherever the site listed its languages by hand. Four places still knew only Arabic, English and Urdu, or only Arabic and English: the redirect that takes a renamed record's old address to its new one (it did nothing under
/fr/and/es/),robots.txt(the rule that keeps crawlers from fetching one copy of the graph explorer per record), the list of changed pages sent to search engines after a production deploy (it also left out Urdu), andllms.txt, which said the museum is published in Arabic and English. The sitemap already listed all five. Search synonyms still exist for Arabic and English only; Urdu, French and Spanish have none yet. - The build works again on the Spanish release branch: the compiled graph had grown too large to be read in one piece. The graph file that every build step reads (
.cache/graph.json) reached 549 MB on 19 November 2026 with five languages of prose, and Node cannot read a file larger than about 512 MB as one text. Validation, the unit tests and the site build all stopped there with the same error. The file is unchanged; it is now read and written a part at a time (scripts/lib/graph-file.mjs), by the site build and by the 20 scripts that used it, and the real file reads back to the same bytes. - French pages no longer stop the site build. The table of other prophets' names had no French entry, so a French page that mentioned the Prophet made the build fail (
/fr/event/hijra/). The French and Spanish tables are both present now. - Spanish prose reads left to right and keeps its section roles. The entity page, the place map and several components treated every language but English as right-to-left (fix taken unchanged from the French lane), and the Spanish headings of the Brief, young-visitors, scholars and reading-level sections are registered. The validator's "Sunni entry first" rule recognises the Spanish label.
- Pages that read the education, kids, tour and card datasets no longer break in French and Spanish. Those datasets exist in Arabic, English and Urdu only; a page such as
/fr/shabab/detective/read a French field that was not there and stopped the build. The loaders now fill a missing French or Spanish field from English, then Arabic, when the data is read, so the page shows the English text until the translation lands. - The Time Browser has its data in French and Spanish. Its generator wrote the per-language files for Arabic, English and Urdu only, which stopped the build at the Time Browser's stop pages.
- The kids' reader is built only where its text exists. A record flagged for the kids' reader got a reader page, and links to it, in every language that had any prose for it; in French that page had no Explorers text and stopped the build, and in Spanish the links led nowhere. The page and its links now exist only where the record's prose in that language has the Explorers section.
- French leveled reading fixed for five records that had two "Pour les Petits" sections (al-dawa-al-sirriyya, abu-sufyan-ibn-harb, abu-talib, abu-ubayda-ibn-al-jarrah, al-aqra-ibn-habis): the first section is kept, the second dropped (still in git history). Six "Pour les Explorateurs" sections (kaaba, battle-of-yamama, abd-allah-ibn-amir, al-bara-ibn-azib, al-qadir-billah, al-qaim-billah) trimmed to 160 words or fewer, removing secondary details only.
Tests
- The site check reads a partly translated strings file.
scripts/check-site.tscrashed on a language whosei18nfile lacks a section (French today); it now falls back to the English string, as the pages do. - The Qur'an passage files are back to their frozen content (branch `release/spanish`). A corpus-wide change on
main(commit7d509c17a5) had inserted direction marks around parentheses in 3,528 of the 6,394 Qur'an passage files: in the Urdu translation text (3,522 values) and in relation notes (21). The Arabic Qur'an text, the search text and the English translation were not touched. On the founder's instruction the 3,528 files are restored to what they were before that commit, and the validator no longer applies the parenthesis rule to the frozen passage files. Until the page renderer isolates them, parentheses in the Urdu translation on passage pages may be drawn the wrong way round. - The accessible books step runs in five languages.
scripts/a11y-books.tskept one list per language for four of the five and crashed on Spanish at the very end of the first full build ofrelease/spanish(after 460,598 pages had been built and checked with 0 errors). French and Spanish now also go through their own braille tables instead of the Arabic path. Known limit: the Spanish and French braille files lose accented letters. - Three tests assert the published languages again.
pipeline.test.ts,tour.test.tsandmodern.test.tsexpected French prose, French tour fields and a four-language fixture that do not exist yet, which failed the build workflow on the release branches. All 634 tests pass.
v2.23.0 — 2026-11-17
Added
- The museum's own podcast feed for "Hudhud's Story Time" (the Kids' Wing story podcast). Two static files under
/podcast/:feed.xml, a complete RSS feed for Season 1 (ten English episodes told by Grandpa and Grandma, each episode linking back to the museum page its story came from, with the show's artist, Kids & Family / Stories for Kids category and owner details the podcast directories require), andhudhud-story-time-cover.jpg, the show's 3000×3000 cover art. The episode audio stays where it is hosted; the feed simply gives the museum a permanent, self-controlled address for Apple Podcasts, Spotify and every other directory, instead of a feed whose show details could not be completed. - Workflow files for the second step of the by-language trial (`main` carries the files only). The by-language build gains a job for the parts of the site that belong to no language and a job that joins the languages' sitemaps, and there are four new deploy workflows: "Deploy shared parts" and "Deploy all parts", each for UAT and for production. GitHub offers a hand-started workflow only when its file is on
main, so the files are here; the code they need is on the Spanish release branch, and until that branch is merged they work only when started on it. The one build and the one deploy are unchanged, and none of the new workflows has run yet. - Workflow files for a trial of building and publishing one language at a time (`main` carries the files only). GitHub offers a hand-started workflow only when its file is on
main, so the twelve files of the trial are added here: "Build islam.museum by language", the shared language deploy, and ten "Deploy<lang>pages" workflows (five production, five UAT). The code they need (the one-language switch in the site build and in the R2 publish) is onrelease/spanishand not onmain, so today they work only when started on that branch, and the production ones cannot work until that code reachesmain. The existing build and deploys are unchanged; nothing was built or deployed with the new files. - French learning activities: all 20 illustrated story decks, 21 source-detective cases and 31 classroom dilemmas now include French, so learners can read the stories, weigh evidence and compare viewpoints in French. Verse references use explanatory paraphrases rather than verse translations.
- French education topic lessons, topics 29–39: eleven of the fifty taught topics (the Great Mosque of Córdoba through the House of Wisdom) now carry French titles, explainers, glossaries and practice questions in
edu/topics-50.yaml, so French readers no longer fall back to English for them. Existing Arabic, English and Urdu text is unchanged; the other topic ranges are translated separately. - French topic lessons: the ten lessons from Medina through Fatima al-Fihri now include French summaries, glossaries and practice questions with answer explanations. Other topic lessons are unchanged.
- French homework search: French classroom phrases now expand to their target labels through French synonym shards and the canon search index, instead of using English-only synonyms.
- French dataset rollout scope: the kids, museum-card and education-support translations are based on the separate French tour/baseline-fixtures work. The large topic lessons and recursive education activities remain untranslated; this partial rollout does not claim full French coverage.
- French education support datasets: age-band introductions, stamp names, seasonal packs, family activities, homework search phrases, location riddles and the approved leveled-reading record labels now include French. The large topic lessons and recursive education activities are not yet translated in this batch.
- French museum-card datasets: all card decks, daily prompts, the thirty-day challenge and editorial word-map titles now include French. Verse-quoting fields use clearly signalled explanatory paraphrases, not newly typed scripture; references, verification notes and age-gating are retained.
- French topic lessons (damascus to hajj-network): twelve of the fifty education topic lessons (Damascus, the rise of Islam, Baghdad and the Silk Roads, al-Andalus, the Mongols and after, the Five Pillars in daily life, Islamic art and architecture, the life of the Prophet, the Qur'an as a text, the Ottomans and gunpowder empires, four cities, and the Hajj network) now include French titles, summaries and glossaries. Qur'an quotations are rendered as explanatory paraphrases, not verse translations; the other topic lessons follow in separate batches.
- French Kids’ Wing datasets: French children’s stories and activities and teacher lesson plans now accompany the existing languages, reducing English fallbacks for French readers. Source references and existing review statuses are unchanged; the wider French dataset rollout is still in progress.
- French object collection (PR #7416). French readers now have full translations of 545 object articles, including their young-visitor and scholarly sections, and French labels and summaries for 549 object records, including four catalogue-only records. One inherited record remains excluded because its Urdu page is missing; no Arabic, English or Urdu pages were changed.
- Ulama Browser: introducing-Islam books, part 2 of category 11 of the 210-book list (8 books). Eight new work records (Armstrong's *Islam: A Short History*, Esposito's *What Everyone Needs to Know about Islam*, Nasr's *The Heart of Islam* and *Islam: Religion, History, and Civilization*, Ramadan, Aslan, Mattson's *The Story of the Qur'an*, Brown's *Muhammad: A Very Short Introduction*), each in Arabic, English and Urdu with author, publisher and year from a catalogue record and linked to the author's page where one exists. Fourteen source records, including the brief's unreadable Memphis and IxTheo links; two Wikipedia sources pinned with ledger records. ⚠ VERIFY: Nasr's *Islam* is dated 2002 (released 24 December 2002; the list gives 2003). Mattson's first-edition year is not settled (catalogue editions from 2007 and 2008), so the record names no year. Also drops five summary clauses from part 1 that no cited source supports (Akbar Ahmed's earlier *Living Islam*, Schimmel's German original, Lang's profession, Denny's audience, Bucaille's profession now "surgeon" as the FUNCI source says).
- Ulama Browser: introducing-Islam books, part 1 of category 11 of the 210-book list (8 books). Seven new work records (Denny, Esposito's *Islam: The Straight Path*, Schimmel, Ruthven, Lang's *Even Angels Ask*, Akbar Ahmed's *Islam Today*, Robinson) and Bucaille's «الكتاب المقدس والقرآن والعلم», upgraded from a draft card; each in Arabic, English and Urdu with author, publisher and year from a catalogue record. Twelve source records, including the brief's unreadable BooksRun and Stanford links; the Wikipedia source is pinned with a ledger record. ⚠ VERIFY: Schimmel dated 1992 (English translation); Akbar Ahmed 1999 (revised edition).
- Ulama Browser: 9 Islamic economics and finance books in «سجلات منتقاة» (category 10 of the 210-book list). Nine new work records (from Chapra's *Towards a Just Monetary System* and Taqi Usmani's *An Introduction to Islamic Finance* to Kuran's *The Long Divergence*), each in Arabic, English and Urdu with author, publisher and year from a catalogue record, plus reviews. Seventeen source records, including the brief's unreadable RelBib and Cambridge front-matter links; the Wikipedia source is pinned to its revision with a ledger record. ⚠ VERIFY: Usmani's book is dated 1998 (first Karachi edition), not 1999; Iqbal and Mirakhor's is dated 2006 (released November 2006; its imprint gives 2007). Also fixes the YAML quoting of one authenticity-ledger record from category 9.
- Ulama Browser books: review fixes for categories 1–8. Thirty-three Wikipedia source records cited by the 210-book records are now pinned to the revision read, each with an authenticity-ledger page record; four books gain their author links (Fahmi Huwaydi ×2, Omid Safi, Sherman Jackson); the coverage register is regenerated from a fresh graph.
- Ulama Browser: 13 family, women and society books in «سجلات منتقاة» (category 9 of the 210-book list). Thirteen new work records (from Mernissi's «الحريم السياسي» and al-Ghazali's «قضايا المرأة» to Abu-Lughod's *Do Muslim Women Need Saving?*), each in Arabic, English and Urdu with author, publisher and year from a catalogue record, plus reviews. Twenty-four new source records, including the brief's unreadable Alexander Street, WorldCat, Everand and ResearchGate links. The ⚠ VERIFY book, Badran's *Feminism in Islam*, is dated 2008 (catalogued release), with its 2009 title-page year noted. Al-Ghazali's book carries no first-edition year; only its 1990 second edition is catalogued.
- Ulama Browser: 24 Islam, modernity, politics and the West books in «سجلات منتقاة» (category 8 of the 210-book list). Twenty-four new work records (from Said's *Orientalism* and Huwaydi's «القرآن والسلطان» to Shahab Ahmed's *What Is Islam?* and Aydin's *The Idea of the Muslim World*), each in Arabic, English and Urdu with author, publisher and year from a catalogue record, plus reviews and bibliographies. Thirty-nine source records, including the brief’s unreadable links. The four ⚠ VERIFY books are settled from catalogues: Al-Azmeh 1993, Devji 2005, March 2009, Aydin 2017. Lewis's *What Went Wrong?* is catalogued from December 2001 and Hallaq's *The Impossible State* from 2014. Feldman's 2008 book is held back: its English title would need a founder-approved exception.
- Ulama Browser: 25 Islamic thought and renewal books in «سجلات منتقاة» (category 7 of the 210-book list). Twenty-five new work records (from Hanafi's «التراث والتجديد» and Fazlur Rahman's *Islam and Modernity* to Abou El Fadl's *Reasoning with God* and Moosa's *What Is a Madrasa?*), each in Arabic, English and Urdu with author, publisher and year from a catalogue, scan, bibliography or encyclopedia record. Thirty-four source records, including the brief's unreadable or mismatched links. The six ⚠ VERIFY books are settled from catalogues: «الصحوة الإسلامية» 1981, al-Turabi's «تجديد الفكر الإسلامي» 1993, Izetbegović 1984 (first English edition), «نحو فقه جديد» 1996, al-Faruqi's *Al-Tawhid* 1982 and *Progressive Muslims* 2003. Al-Sharafi's «الإسلام بين الرسالة والتاريخ» is dated 2001, not 2008, and *Reasoning with God* was published by Rowman & Littlefield, not Cambridge.
- Ulama Browser: 13 Sufism and spirituality books in «سجلات منتقاة» (category 6 of the 210-book list). Thirteen new work records (from Ernst's *Words of Ecstasy in Sufism* and Chodkiewicz's *Le Sceau des saints* to Knysh's *Sufism: A New History*, «لأنك الله», «فاتتني صلاة» and *Secrets of Divine Love*), each in Arabic, English and Urdu with author, publisher and year from a catalogue, review or reference record. Twenty-two source records, including the brief's unreadable LIBRIS, Free Library and Biblio links.
- Ulama Browser: 16 creed, kalam and philosophy books in «سجلات منتقاة» (category 5 of the 210-book list). Sixteen new work records (from al-Attas's *Islam and Secularism* and Nasr's *Knowledge and the Sacred* to al-Jabri's «بنية العقل العربي», Taha Abderrahmane's «روح الحداثة» and Adamson's *Philosophy in the Islamic World*), each in Arabic, English and Urdu with author, publisher and year from a catalogue, scan or encyclopedia record. Twenty-two source records, including the brief's unreadable PhilPapers link. Of the two ⚠ VERIFY books, Ali Harb's «نقد النص» is dated 1993 from his Wikipedia bibliography; Hanafi's «مقدمة في علم الاستغراب» carries no first-edition year because sources disagree (1988–1992). Leaman's book first appeared in 1985 as *An Introduction to Medieval Islamic Philosophy*.
- Ulama Browser: fiqh, legal theory and maqasid books in «سجلات منتقاة» (category 4 of the 210-book list). Eighteen new work records (from al-Zuhayli's «الفقه الإسلامي وأدلته» and Kamali's *Principles of Islamic Jurisprudence* to Hallaq's *Shari'a* and al-Raysuni's «مدخل إلى مقاصد الشريعة»), each in Arabic, English and Urdu with author, publisher and edition from a catalogue, scan or publisher record; the Kuwaiti Fiqh Encyclopedia joined the browser. Twenty-four source records, including the brief's unreadable Academia and HathiTrust links. Four ⚠ VERIFY books (al-Alwani, «في فقه الأولويات», «الفقه المنهجي», «الفكر المقاصدي») carry no first-edition year, only the edition a source shows; two books are held back for want of a readable source (al-Raysuni's «نظرية المقاصد عند الإمام الشاطبي» and al-Qaradawi's «الاجتهاد المعاصر»). Hallaq's *An Introduction to Islamic Law* is dated 2008, not 2009.
- Ulama Browser: 21 sira and modern Islamic history books in «سجلات منتقاة» (category 3 of the 210-book list). Sixteen new work records (from Lings's *Muhammad* and Maalouf's «الحروب الصليبية كما رآها العرب» to Hoyland's *In God's Path*), each in Arabic, English and Urdu with author, first-edition year and publisher from a catalogue, scan or publisher record; five existing records joined the browser (al-Nadwi's and al-Sallabi's «السيرة النبوية», «الرحيق المختوم», al-Umari's «السيرة النبوية الصحيحة», and Donner's *Narratives of Islamic Origins*, upgraded from a draft card). Thirty-two source records, including the brief's unreadable links. Al-Umari's first-edition year (the list's ⚠ VERIFY) is still unconfirmed, so his record states only the 1994 sixth edition; *In God's Path* is dated 2014, not 2015.
- Ulama Browser: 16 hadith and Sunnah books in «سجلات منتقاة» (category 2 of the 210-book list). Fifteen new work records (from «تيسير مصطلح الحديث» and Azami's *Studies in Hadith Methodology* to Brown's *Misquoting Muhammad*), each in Arabic, English and Urdu with author, first-edition year and publisher from a catalogue, scan or review; Juynboll's *Encyclopedia of Canonical Hadith* moved from a draft card to a sourced record and joined the browser. Twenty-four source records, including the brief's unreadable links, each saying what was and was not read. The list's ⚠ VERIFY on al-Qaradawi's «كيف نتعامل مع السنة النبوية» is settled: first edition 1990 (Dar al-Wafa and IIIT). «منهج النقد في علوم الحديث» is catalogued from 1970, before the list's 1979.
- Ulama Browser: 21 modern Qur'an and tafsir books in «سجلات منتقاة» (category 1 of the 210-book list). Eighteen new work records (from «صفوة التفاسير» and «التفسير المنير» to Neuwirth's *The Qur'an and Late Antiquity*), each in Arabic, English and Urdu with author, first-edition year and publisher from a catalogue, publisher or library record; three existing records (Ibn Ashur's «التحرير والتنوير», the Encyclopaedia of the Qur'an and the Sha'rawi tafsir) joined the browser. Thirty source records, including the brief's unreadable links, each saying what was and was not read. Two dates differ from the list: «صفوة التفاسير» first appeared in 1980 (not 1978), and the Open Library records date Abdullah Saeed's two books to 2005 and 2013.
- Heba Kotb's remaining brief sources recorded (person/heba-kotb). Eight source records for the links in the 18 November brief that the enrichment had not yet recorded (Jarir, Kobo, Audible, her Arabic About page, The Arabist, the YouTube clip and Matn News); each says what was and was not read. No prose changes.
- Youssef Ziedan's record enriched (person/youssef-ziedan). Novel dates corrected («النبطي» 2010, «محال» 2012, «جونتنامو» 2014, «نور» 2016); book count now more than seventy; a new section on «سفر العذارى» (Dar N, 2025) and the January 2026 Book Fair dispute; Azazeel's translations from IPAF's list; dated Kuwait Foundation (1994, 2006) and Shuman (1996) prizes; further titles; in Arabic, English and Urdu. His faith entry is untouched.
- Islam Behery's record enriched (person/islam-behery). The 2015 case told in full: the in-absentia sentence, the October 6 acquittal on the same charges, the July 2016 Cassation ruling, CPJ's and Amnesty's responses and the show's halt in April 2015; the pardon committee's recommendation; the October 2024 cheque reconciliations and the withdrawn namesake complaint; and a new section on his case in scholarship; in Arabic, English and Urdu. His faith entry is untouched.
- Khaled Montaser's record enriched (person/khaled-montaser). His novel «زمن سعاد» and his May 2026 book on the Diaa El-Awady phenomenon, the 2026 Reesha edition of «وهم الإعجاز العلمي», his first book dated 2003, more titles with publishers and ISBNs, his column in Al-Watan, later honours as the press reported them, the 2018 family statement, and the Kharej Al-Nass episode on Farag Foda reworded (Foda was killed in 1992); every brief link recorded as a source; in Arabic, English and Urdu. His faith entry is untouched.
- Saad El-Din El-Hilali's record enriched (person/saad-eldin-el-hilali). His Saudi secondments of 1988–1993 and further committee memberships, his February 2014 speech likening two ministers to Moses and Aaron as the press reported it and Al-Azhar's response, Al-Azhar University's own disavowal of April 2025 kept distinct from the Fatwa Centre's, the precedents he cited for his inheritance reading, his October 2025 segment on shrines after the al-Badawi mawlid, and a refreshed, dated Facebook count; in Arabic, English and Urdu. His faith entry is untouched.
- Heba Kotb's record enriched (person/heba-kotb). Her later full professorship, the «للكبار فقط» series read as a series, both accounts of «بيت العز», the 2002 and 2004 column dates kept apart, a new Honours and recognition section from her CV and site with the four honours an independent directory confirms, her 2010 Oprah appearance and her reaction to it, more programmes, books and institution work, each attributed and cited, in Arabic, English and Urdu. Her faith entry is untouched.
- Heroes of Islam Explorer: قاعة الأبطال (`/fr/heroes/`). A new front door over the one graph: six virtue halls — Sacrifice, Courage, Mercy, Generosity, Knowledge and Steadfastness — each a
data/exhibittrail (exhibit/hero-*) whose four rooms name a hero person and the stand they took, backed by six long-form trilingual narratives (content/{ar,en,ur}/exhibit/hero-*.md) and enriched "heroic stand" sections on five thin hero records (Fatima al-Fihri, Sumayya, Saladin, Abd al-Qadir al-Jaza'iri, Rabi'a al-Adawiyya) in all three languages. The page is build-rendered with no client script: oversized display-type names, an eight-fold starVirtueMedallionper hall in its own accent — grandeur by calligraphy and geometry, no figurative imagery of any person./api/v1/heroes.jsonexports the roster; the Explore menu entry and the landing card ship in Arabic, English and Urdu. Plandocs/plan/71-heroes-explorer.md(approved 17 November 2026, non-impacting). Built by Devin (SWE-2 Medium). - Ulama Browser: متصفح المفكرون والعلماء والأدباء والمحدثون والكتاب (`/fr/ulama/`). A new browser over the one graph listing every thinker, scholar, litterateur, muhaddith and writer the museum knows — the founder's five categories plus all knowledge professions (scholars, muhaddiths, jurists, muftis, qadis, theologians, philosophers, historians, geographers, writers, teachers, preachers, scientists, mathematicians, astronomers, physicians, calligraphers; poets keep their own wing) — and everything the Qur'an, the Sunnah, the seerah and every harvested book says about them. Persons qualify at build time by role or occupation job in
scripts/ulama.ts; mentions join three ways: canonicalmentionsedges, a reviewed Qur'an verse table (Luqman the sage, the people-of-knowledge verses, the verses of the pen and of writing — the frozen Tanzil text is never searched), and a guarded folded-spelling name lexicon in which a bare common name never counts alone, so a book harvested later brings its scholars and its mentions automatically. Sections: the four passage corpora plus five people sections (ulama, muhaddithun, mufakkirun, udaba, kuttab) and a curated shelf; category chips alim/muhaddith/mufakkir/adib/katib/scientist. The signature view is «مجلس العلم» — five ink-pillars sized by the records they hold under the majlis arch, with the most-mentioned figures as live links./api/v1/ulama.json, an Explore-menu entry and a landing card ship in all three languages; the plan isdocs/plan/70-ulama-browser.mdand the acceptance specdocs/engineering/62-ulama-browser-requirements.mdkeeps the founder's original brief — and his two addenda naming six modern figures and asking for full-detail proses — verbatim for production validation. Built by Devin (SWE-2 Medium). - Munazarat Browser: متصفح المناظرات بين الاسلام والأديان الأخرى حجة بالحجة ودليل بالدليل وتاريخ بالتاريخ وكلمة بالكلمة (`/fr/munazarat/`). A new browser over the one graph listing every debate and disputation the museum's sources carry — between Islam and the other religions (the mubahala of Najran, Ja'far at the Negus's court, the Qur'anic disputations with the scriptuaries and the polytheists) and inside the ummah (Ibn Abbas and the Khawarij, the arbitration of Siffin, the Mihna interrogations, Umar II's debate) — in the Qur'an, the Sunnah, the seerah, Companion and Tabi'in reports, and every harvested book, present and future. Passages are found at build time by a guarded lexicon in
scripts/munazarat.ts(مناظرة nouns and مناظر verbs, جدال/محاجّة/احتجاج/خصومة, مباهلة/نبتهل anchors; the li'an oath, حَجّ pilgrimage and braid senses excluded) so a book harvested later lands automatically. Records group into five corpus sections (quran, sunnah, seerah, books, curated) with debate/mention kinds and scriptuaries/polytheists/intra/other opponent chips. Eleven curated records: three new events with full trilingual prose (mubahala-of-najran,ibn-abbas-khawarij-debate,jafar-debate-at-najashi-court) citing ibn-hisham-sira, sahih-muslim, musnad-ahmad and al-bidaya-wal-nihaya, plus eight existing events tagged in place. The signature view is «حلبة المناظرة» — a pure-SVG debate floor flowing opponent → kind → corpus, keyboard-reachable and reduced-motion safe./api/v1/munazarat.json, an Explore-menu entry and a landing card ship in all three languages; the plan isdocs/plan/69-munazarat-browser.mdand the acceptance specdocs/engineering/61-munazarat-browser-requirements.md, which keeps the founder's original brief verbatim for production validation. Built by Devin (SWE-2 Medium). - Sport Browser — Sports and Competitions in Islam (متصفح الفروسية والسباحة والرماية وجميع رياضات ومنافسات المسلمون, `/fr/sport/`). New browser over the one graph listing every sport and contest mention — horsemanship, swimming, archery, wrestling, races and wagered contests, spear-play, ball and board games, hunting, modern sport — in the Qur'an, Sunnah, seerah, Companion and Tabi'in reports and every harvested book. Detector
scripts/sport.ts: guarded lexicon in eight families (khayl/ramy/sibaha/sira/sabq/laib/sayd/riyada), context guards on ambiguous words, id-based false-positive exclusions; future harvested books join automatically. Newsportclass (kind/era/status_today/furusiyya) with 17 seeds, 5 contest event records, 10 existing records taggedsport-browserin place; athletes read fromperson.sports— no side list. Signature viewsig/SportFields.astro(the contest ground);/api/v1/sport.jsonwith section/topics/signals; Explore entry, landing card, trilingual ar/en/ur pages and prose; plandocs/plan/66-sport-browser.md, specdocs/engineering/58-sport-browser-requirements.md(SPORT-R1..R11, original requirement verbatim), lanesport. - Umuma Browser — Motherhood in Islam (متصفح الأمومة في الإسلام, `/fr/umuma/`). A new browser over the one graph listing everything the museum's sources mention about motherhood — the mother and her standing, pregnancy, childbirth, nursing and milk-kinship, the custody and raising of children, and dutifulness to parents — in the Qur'an, the two Sahihs, the Muwatta and the Sunan, the seerah books, Companion and Tabi'i reports and every harvested book, present and future. Passages are found at build time by a guarded lexicon in
scripts/umuma.tsacross six topical families (umm/haml/wilada/radaa/tarbiya/birr): bare «أم» counts only beside family vocabulary («أم القرى», «أم الكتاب» and «أمة» excluded), «حمل» counts in the pregnancy sense only beside pregnancy vocabulary (carrying/transport excluded), «ولد» the boy is out and only the birth forms count, «بر» counts only beside parent vocabulary, and verified false positives drop out by id — so a book harvested later brings its motherhood mentions automatically. Records are grouped into five corpus sections (quran, sunnah, seerah, books, curated) with topic chips; mothers on record surface the graph's ownchild_ofedges to female parents — no side list. Twenty existing records taggedumuma-browserin place (umm, ummahat-al-muminin, rada, rada-suckling-fiqh, istilad-umm-al-walad, birr-al-walidayn, aqiqa, aqiqah, kafalat-al-yatim, tahnik, and the mothers Amina bint Wahb, Maryam, Hajar, Khadija, Fatima, Sumayya bint Khayyat, Halima al-Sa'diyya, Asiya and Fatima bint Asad) plus five new citedconceptseeds —umuma,haml,wilada,hidana,birr-al-umm— each with trilingual label, summary and prose cited to the Tanzil Qur'an, the two Sahihs, the Muwatta, Sunan Abi Dawud and the Kuwaiti Fiqh Encyclopaedia. Signature viewsig/UmumaLines.astro: the six lines of motherhood, a disc per corpus section on each line sized by record count./api/v1/umuma.json, an Explore-menu entry and landing card, trilingual ar/en/ur pages, plandocs/plan/64-umuma-browser.md, acceptance specdocs/engineering/56-umuma-browser-requirements.md(UMUMA-R.., original Arabic/English requirement recorded verbatim incl. the microSD/merge instructions), laneumuma. Records are draft until a human reads them. - Umuma Browser enrichment wave 1 (the plan's scoped next wave): three new cited
conceptseeds —nafaqat-al-murdia(maintenance of the nursing mother, Q 2:233),iddat-al-hamil(the pregnant woman's waiting period, Q 65:4) andkhala-min-al-radaa(the maternal milk-aunt, Q 4:23 + the "what is forbidden by suckling" report) — each with trilingual label, summary and prose cited to the Tanzil Qur'an, the two Sahihs and the Kuwaiti Fiqh Encyclopaedia, taggedumuma-browser; six more mothers tagged in place (Umm Ayman the Prophet's foster-mother, Umm Sulaym mother of Anas, Umm Ruman mother of Aisha, Salma Umm al-Khayr mother of Abu Bakr, Umm Haram bint Milhan, Safiyya bint Abd al-Muttalib mother of al-Zubayr). - The Bunuwwa Browser (متصفح البنوة في الإسلام · Sonhood and Daughterhood in Islam): a new front door listing everything the sources mention about sons, daughters, offspring, kindness and disobedience to parents, orphans and suckling, the rights of the child and the named children of the stories — Qur'an verses, hadith reports, seerah books and every harvested work, detected at build time by a guarded lexicon so future books join automatically. It carries a signature view (six ruled lines of the filial topics), a famous-parents-and-children section read from the graph's child_of edges, sixteen tagged-in-place records and nineteen person records plus four new cited concepts (البنوة، الرضاعة، الموءودة، الذرية) with trilingual prose, an Explore-menu entry, a landing card and the
/api/v1/bunuwwa.jsonexport. - Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episode 28 added, the second of the Aisha bint Abi Bakr episodes — the recap of the series plan and the standing rule verse; the wives' list read off screen in order with the nine-rooms explanation and the book list plus two new author names; the speaker's and audience's age-question answers repeated from episode 27 and kept attributed and under perspectives; the two families' late departure from Mecca in the hijra; the wedding night in Shawwal of year two AH after Badr and the 'laylat al-bina' naming; the Shawwal joke and the consult-the-scholars rule; the wedding day as Aisha narrates it (the swings, the Ansar house, the walima milk and the shy girls); the simplicity of the house (the fibre-stuffed cushion, the water skin, the mat, the sayings of al-Hasan and Sa'id ibn al-Musayyab); the proofs of love from Masruq, Ammar and Amr ibn al-As's 'who is dearest' list; the knot-in-the-rope hadith; the gifts-on-her-day story with Umm Salama and Fatima's missions and Fatima's later secret; the audience comments on the age gap and on equality in love with the 'this is my division' supplication; the outings (the Abyssinians' spears in the mosque, the two races, the 'O A'ish' endearment); the menstruation-cup tenderness; the broken-dish story attributed to the two Sahihs on the speaker's word; the arbitration story with Abu Bakr; the 'Lord of Muhammad / Lord of Ibrahim' oath hadith; and the closing lesson that a house built on God's obedience lasts; in Arabic, English and Urdu drafts, each statement cited to the recording by timestamp, the contested age and ifk-adjacent material placed under perspectives and never asserted by the museum, audience statements attributed to their sayers, and the museum's readings of garbled names and the repetition loops listed in the 'left out' sections; machine transcript, not listened to.
- Twenty-two more records of Ibn Sina's Canon, Book III (the diseases of each organ, head to foot), volume 2, printed pages 350 to 370 (pages.json indices 349 to 369): the rest of the tenth fann's fourth maqala — peripneumonia (the hot tumour of the lung), its signs and its passage into suppuration; the lung's hard, soft, pustule and water gatherings and the windpipe's tumour and wound; the pus and the filling of the chest cavity, its causes, its four outcomes, its side-signs and the sound and bad bursting, with the author's and Galen's witnessings on the amounts spat; the ulcers of chest, diaphragm, lung and windpipe — the lung ulcer being consumption — with the physicians' disagreement and Galen's ruling on what heals; the causes of the lung's ulcers; those predisposed to consumption in build, complexion, age, country and temperament and what they must shun; and the consumption's signs to the end of the maqala — then the whole of maqala 5, the treatments: the common matters (the ordered venesections, the cupping and scarification, Galen's warning on purging under severe fever, the regimen of drinks and waters by season); the pleurisy treatment (the evacuations by blood and by humour, the direction of the pain ruling venesection or purge, the feeding ladder from barley water) and its poultices, liniments, moist and dry compresses, cupping glasses and the ancients' lohochs and under-tongue pills; the non-pure phlegmatic and melancholic kinds and the water-in-lung treatment; peripneumonia's treatment; the discourse on suppuration (aiding ripening, aiding the burst, the post-burst cleansing, the fine cautery and the lancet drainage, all carried as the book's descriptions); and the chest-region ulcers' treatment — the true cure and the palliation — with the milks (women's, donkey's, goat's, mare's, cow's), the accomplished physician's milk-drinking method, the buttermilk method, and the consumptive's foods. Each is a trilingual concept in the Medicine Browser cited to the printed page; recipes, doses, drug and plant names, decoctions, electuaries, pills, lohochs, draughts, juices, oils, salves, cerates, poultices, fumigations, clysters, suppositories and food prescriptions are not reproduced; the animal- and human-derived lines (women's milk suckled from the breast, donkey, goat, mare, cow and sheep milks, the fox's lung, the cow and goat kidney fats, the mountain-cow dung, the tar spoonful, the trotters, the plucked-claw crabs) are kept as labelled listings for the founder; the toxic and stupefying lines (the arsenic fumigations, the stupefying admixtures, the poppy-descended syrup and the under-tongue pills) are named only as kinds; the sayings and reports (Galen, Hippocrates, Yusuf al-Sahir, the accomplished physicians' milk method, the ancients' vs the moderns' practice, the author's own witnessings and the tried-remedy lines) are carried as the book's attributions and labelled; and the hand-works (the red-clay cloth pus-finder, the fine cautery piercing the chest, the lancet drainage, the pus-drawing instrument, the uvula's cutting, the postures, the chair-shaking, the heated-stone milk cooking, the hot iron in the buttermilk) are carried as the book's descriptions, not as procedures.
- Wa Nalqa al-Ahibba, Amr Khaled (lecture series, Assabile archive): episode 27 added, the first of the Aisha bint Abi Bakr episodes — the introduction to Aisha as the speaker presents her: the repeated rule verse, the wives' quiz and the book list with a new al-Bouti title; her place as the Prophet's beloved, the only virgin he married on the speaker's word, daughter of Abu Bakr; the house of Abu Bakr (Qutayla mother of Asma and Abdullah, Umm Ruman mother of Aisha and Abd al-Rahman, the half-sisters); Umm Ruman's emigration and death after the ifk incident and the Prophet praying over her and descending into her grave; Abd al-Rahman ibn Abi Bakr's late conversion after fighting at Badr and Uhud; the screen's numbers (born year five of the mission, betrothed at seven, married at nine, widowed at eighteen, dead on 17 Ramadan 57 AH aged sixty-six, the transcript contradicting itself at points); Khawla bint Hakim's visit to Umm Ruman; Abu Bakr's hesitation because al-Mut'im ibn Adi's son Jubayr had asked for her; the lesson on keeping one's word and on steady character toward the disbeliever; al-Mut'im's wife speaking first and releasing Abu Bakr; the Prophet's betrothal; the audience's mother-in-law comment; the woman's question about the nine-year age and the whole answer session (early maturity in hot climates, environment and custom, the earlier betrothal, examples of Hafsa, Safiyya and Amina, the doubt's recentness, a quoted English Orientalist via Aisha Abd al-Rahman's book, the speaker's own caution that girls today must not imitate the age); the why-he-married-her section (a young long-lived transmitter of the Prophet's private life, her narrated questions and stories, the four caliphs' marriage ties, Muhammad the man versus Muhammad the messenger, and the dream hadith he attributes to al-Bukhari); in Arabic, English and Urdu drafts, each statement cited to the recording by timestamp, the speaker's and audience's contested answers about her age placed under perspectives and never asserted by the museum, the audience's statements attributed to their speakers, and the museum's readings of garbled names and the repetition loops listed in the 'left out' sections; machine transcript, not listened to.
- Job Browser linking wave 49: 37
occupations:edges on 25 person records — finishes the astronomer pool (al-Battani, Habash, Jabir ibn Aflah) plus fifteen imams, eight mathematicians, and geographer, physician, preacher and alim links. - Wagdy Ghoneim's «الأخلاق الحميدة» series on the Assabile archive is now covered in full: lessons 3 (truthfulness, sidq, in its four kinds — intention, faith, deed and speech) and 4 (fidelity, wafa', to God, people and oneself across covenants, vows and daily conduct), each statement cited to the recording by timestamp, in Arabic, English and Urdu.
- Wagdy Ghoneim (Egyptian preacher) joins the museum with his Assabile archive, starting with the first two lessons of «الأخلاق الحميدة» (noble ethics): lesson 1 on trustworthiness (amana) and its kinds, lesson 2 on sincerity (ikhlas), each statement cited to the recording by timestamp, in Arabic, English and Urdu.
- Heba Kotb (person/heba-kotb): a new published record for the Egyptian physician and broadcaster — professor of forensic medicine at Cairo University and Egypt's first licensed sexologist, presenter of «كلام كبير» (The Big Talk) since 2006 — with full trilingual prose (a long-form biography: her forensic-medicine and Florida sexology doctorates, her clinic and advice columns, her television career, her book «للكبار فقط», her online courses platform, the international coverage from AP to Reuters, and the public debate around her work carried as attributed claims) and ten new source records (her Wikipedia articles in Arabic and English, her official site and CV, AP 2006, ABC News 2007, Arab Media Society 2007, Reuters 2021, her agency roster and a 2025 newsroom profile). Researched and written at the founder's direction on 17 November 2026.
- Saad El-Din El-Hilali (person/saad-eldin-el-hilali): a new published record for the Al-Azhar professor of comparative fiqh and television scholar — the standing religious-segment holder on Amr Adib's «الحكاية» (MBC Masr) since 2018 — with full trilingual prose (a long-form biography: his Mansoura origins and Al-Azhar career to full professor in 1996, the Kuwait University years, his verified books from «الإسلام وإنسانية الدولة» to the seven-volume funeral-fiqh encyclopaedia, his stated method of presenting every fiqh opinion without concealment, and the scholarly reception — Ali Gomaa's review, Al-Azhar's Global Fatwa Centre rejection of April 2025, Dar al-Ifta responses — carried as attributed claims) and fifteen new source records (his Arabic Wikipedia article, scholar profiles, his official Facebook page, MBC's segment pages, a books catalogue and dated Egyptian press). Researched and written at the founder's direction on 17 November 2026.
- Khaled Montaser (person/khaled-montaser): a new published record for the Egyptian physician, columnist and broadcaster — Cairo medicine class of 1983, dermatology and venereology specialist, presenter of «خارج النص» on ON TV from 2015 and a long-standing Al-Masry Al-Youm columnist — with full trilingual prose (a long-form biography: his Giza birth and Damietta upbringing, the Suez Canal Authority department headship, his newspaper career and 2010 Bahrain columnist prize, his television path, his books from «وهم الإعجاز العلمي» to the 2024 memoir «خلف خطوط الذاكرة», his anti-pseudoscience campaigns and his public project in his own framing, and the reception including Dar Al-Iftaa's conditional fatwa of 2011, carried as attributed claims) and fifteen new source records (his Arabic Wikipedia articles, Wikidata, interviews, Dar Al-Iftaa via Islam Story, his Facebook page and dated audience snapshots). Researched and written at the founder's direction on 17 November 2026.
- Wagdy Ghoneim's «الإیمانیات» series on the Assabile archive is now covered in part: episodes 1-7 of 13 (faith in God and its three foundations and its bearing on righteous action; faith in the revealed books and the Qur'an as their seal; faith in the messengers, their humanity and their call to tawhid), each statement cited to the recording by timestamp, in Arabic, English and Urdu; episodes 8-13 are not yet mined.
- Islam Behery (person/islam-behery): the record landed earlier today is enriched — full trilingual prose replaces the stub (English about 1,970 words), 17 further cited sources added (19 in all), birth year 1974 recorded with the source discrepancy noted, a lived_in edge to Cairo and his full-name variants added; the faith line as repaired on main is kept.
- Youssef Ziedan (person/youssef-ziedan): his published record is enriched with full trilingual prose — the English page grows from about 370 words to about 3,000 (Arabic about 2,730, Urdu about 3,230) — covering his Sohag birth and Alexandria upbringing, the Bibliotheca Alexandrina Manuscript Center years and the dispute that ended them, his scholarship on Sufism and Arabic medicine, the novels in order through «الحصيد» (2025) with a full section on «عزازيل» and its reception, and his co-founding of the Takween Foundation in May 2024 and resignation seven weeks later; three new source records added, the record's faith line untouched.
- Wagdy Ghoneim's «الإیمانیات» series on the Assabile archive is now fully covered: all 13 episodes (faith in God and its foundations and its bearing on righteous action; faith in the revealed books and the Qur'an as their seal; faith in the messengers, their humanity and their call to tawhid; faith in the angels and their works in the unseen and the witnessed; faith in the Last Day, its signs and its reckoning of deeds; faith in divine decree in its legal and creative senses), each statement cited to the recording by timestamp, in Arabic, English and Urdu.
- Wagdy Ghoneim's «غزوات النبي» series on the Assabile archive is now covered: all 8 episodes (introduction and Badr; the events and lessons of Badr; the Treaty of Hudaybiyya; the Battle of the Confederates or the Trench; the campaign of Khaybar; the campaign of Tabuk; the battle of Hunayn; the conquest of Mecca), each statement cited to the recording by timestamp, in Arabic, English and Urdu.
- Wagdy Ghoneim's 'Sharh Mabadi' al-Tawhid' (a commentary on the principles of tawhid) series on the Assabile archive is now covered: all 12 episodes (the three divisions of tawhid — essence, names and attributes, lordship, divinity; the shirk of following desires; taking allies other than God; detailed manifestations of shirk — calling on other than God, intercession, vows, means of approach; the attributes exclusive to divinity; the worship of the dead and natural phenomena; Satan of the jinn and his war on the children of Adam; and the closing episode on the two testimonies and proofs of divinity), each statement cited to the recording by timestamp, in Arabic, English and Urdu.
- Wagdy Ghoneim's 'Solok al-Muslim al-Multazim' (conduct of the committed Muslim) series on the Assabile archive is now fully covered: all 26 episodes (akhlaq and the meaning of commitment; appearance and cleanliness; daily adhkar and guarding the tongue; eating, dressing, sneezing and greeting etiquette; parents living and dead; neighbour and driving ethics; faith as certitude, confession and deed; sincerity and conformity; the unseen and the witnessed), each statement cited to the recording by timestamp, in Arabic, English and Urdu; episodes 14-26 cover the heart's works — repentance, reliance, gratitude, patience — the names of God, and the closing episode on the Last Day; each statement cited to the recording by timestamp.
- The French Qur’an import tool recognizes Kazimirski’s public-domain translation. Passage pages still need a verified verse-by-verse alignment before they can display it.
Changed
- The virtual tour now includes French exhibit labels, gallery descriptions and journey narration.
- Ten teen learning topics, from Al-Khwarizmi to Ibn Battuta, now offer French explainers, glossaries and practice quizzes.
- French translations of event records from arrest-of-al-khaqani-vizier to daraqutni-rihla-egypt-350: prose pages in
content/fr/event/and French labels and summaries on each record, translated from English with their citations and links kept, added packet by packet; the French translation queue (scripts/fr-queue.ts) no longer crashes on start. - Seven Futuh al-buldan event records (asad-khurasan-rule, asawira-basra-settlement, asawira-defection, ashras-jizya-episode, ata-inheritance-cases, basra-bathhouse-permissions, basra-crowd-abdallah-ibn-ali) now record their source in provenance, taken from the record's own source list; the Arabic date notes of the Kleber assassination record no longer use literal ellipses inside quotations.
- The public-claims ledger records the French release’s bibliography counts after registering Kazimirski, separately from live-site figures.
- French interface, thematic collections: the French edition (
/fr/) now shows its own wording for the thematic browsers, from the Fatwa Explorer through Heroes of Islam (about sixty sections, some 3,200 strings ini18n/fr.yaml) — explorers of the unseen, worship, ethics, family, the prophets, the Qur'an sciences, scholars, battles, careers and courses — instead of falling back to English. Machine translation from the English pivot, placeholders and links unchanged.
Changed
- The virtual tour now includes French exhibit labels, gallery descriptions and journey narration. The French Farewell Pilgrimage description specifies that it was the Prophet’s only hajj.
Fixed
- French reading and navigation (PR #7416). French pages read left to right; the reading-level headings and brief, young-visitor and scholarly sections are recognized in French. The translation queue also finds Arabic-only sacred records. Object translations were checked against the museum’s Arabic prose and catalogue data to correct inherited English-pivot errors without silently rewriting Arabic-agreeing source conflicts.
- French kids' reader no longer stops the site build. The leveled-reading headings now have their French forms (“Pour les Petits”, “Pour les Explorateurs”, “Pour les Chercheurs”) in
edu/levels.yaml, so a French article written with them is recognised by the validator and by the kids' reader (/fr/sighar/read/…). Before this, the first French record flagged for children (building/bayt-al-hikma) made every pull request to the French release branch fail at the build step. - French Time Browser links no longer stop the site build. The Time Browser's prepared record lists are now written for French as well as Arabic, English and Urdu, with French titles and summaries where a record has them, so the old per-stop addresses (
/fr/time-browser/…/…/) redirect to the museum entry like the other languages. That redirect page also now declares French as left-to-right; it used to mark every language except English as right-to-left. - The after-build site check reads a partly translated language the way its pages do.
scripts/check-site.tsnow takes its strings through the site's own loader, which fills a key French has not translated yet from English, so the check compares each page with the text it really shows instead of crashing on the missing French section. The embed index and not-found page checks are unchanged for Arabic, English and Urdu. - French prose perspective labels such as “Point de vue sunnite” now satisfy the Sunni-first check, while negated forms such as “non-sunnite” do not. Validation and rendering share label parsing and accept regular spaces, non-breaking spaces and narrow non-breaking spaces before the colon. French community names and attribution verbs also participate in the existing label-before-view warning.
- French release verification: missing French dataset fields now have safe tour, gallery and world-page fallbacks, French prophet-name guards keep honorific rendering scoped correctly, and localized component types accept optional French fields during the rollout. The French queue imports the existing repository path helper correctly. Validate now builds the graph cache before its unit tests; regression tests cover French routes, search-index routing and the R2 sync prefix configuration.
- A web address in Arabic or Urdu text no longer swallows the punctuation after it. A bare address typed in a summary or article (
www.example.orgorhttps://…) now ends before «؛», «،», «؟» and «۔», the way it already ended before a Latin full stop, comma or semicolon. Before this, «…والموقع www.semerkand.tv؛» made the semicolon part of the link, so the link on the source pagewikipedia-semerkand-tv-diqwas broken and the full-site link check onmainfailed; the IslamQA listing page had the same fault with «،». Bracketed links are unchanged. - Arabic and Urdu parentheses now display the right way round. About 20,400 files had plain brackets around Arabic-script text, which right-to-left pages draw turned the wrong way, such as «الخطيب (المؤلف)»: 247 Arabic and 7,052 Urdu articles, the Arabic and Urdu UI strings and about 13,100 data records. They now carry the U+2066/U+2069 isolates the rule requires, and
npm run validaterejects a plain one in new Arabic or Urdu content. The Arabic and Urdu “approximately” date tag and the names in the art and Muslim-world data files of the site got the same fix. - ﷺ now appears only for Prophet Muhammad, in Arabic, English and Urdu. The page renderer used to add it to the first «النبي» / "the Prophet" / «رسول اللہ» of every article, so «معركة النبي صموئيل», "the Prophet Daniel", «عيسى رسول الله» and a bare «النبي» on another prophet's own page carried Muhammad's mark (about fifty such marks across the site, found by rendering every article before and after). A title with another prophet named beside it is now left unmarked; the rule for new content is in
AGENTS.md, andnpm run validatechecks that source texts keep ﷺ only beside Muhammad.
Tests
- The reach-index test counts French Wikipedia, the graph test checks a record that already has French prose, and CI builds the graph cache before running tests.
What comes next
- Occupations woven into the graph: the jobs taxonomy is served as records today; next it becomes connections, so a person links to their craft and a question like “every calligrapher of the ninth century” can be walked across the graph.
- Query by Wikidata ID: a published index mapping all 16,232 Wikidata identifiers in the data to their records, with a lookup address for each QID.
- Formal API documentation (OpenAPI) and library-standard exports: OAI-PMH, MARC and a Dublin Core mapping.
- The unharvested corpora — the DİA encyclopaedia, the women hadith scholars of Nadwi's 57-volume dictionary, the narrator (rijal) graph — and new language editions: Indonesian, Turkish and French.
The complete history, commit by commit, lives in the public repository (github.com/amado2k5/islam-museum); this page is its reader's digest.