angryreply/jawami3_alkalem

جوامع الكلم — Jawami' al-Kalim hadith database

The complete data of Jawami' al-Kalim 4.5 (1431 AH), the hadith program created by Offok and published free of charge by Islamweb as a waqf, extracted from its Windows-only binary format into a single SQLite file.

• 1,400 musnad books and 824,841 narrations, full Arabic text with tashkeel
• 638,237 isnads as ordered narrator chains, linked to 49,845 narrator biographies with their rank and 259,698 critics' verdicts
• Gradings, and takhreej clusters that group every narration of a hadith across all books
• Subject trees, the Quran, six classical Arabic dictionaries, and four major shuruh
• A ready-made full-text search index (SQLite FTS5)

Browse it at https://hadith.adhamzahran.com. The extraction code and the site are at https://gitlab.com/lordadamson/hadith.adhamzahran.com (MIT).

Download

The file is jawami.db.zst, a zstd-compressed SQLite database: a 480 MB download that decompresses to 4.2 GB.

hf download angryreply/jawami3_alkalem jawami.db.zst --repo-type dataset --local-dir .
# or: curl -LO https://huggingface.co/datasets/angryreply/jawami3_alkalem/resolve/main/jawami.db.zst

zstd -d jawami.db.zst

On Windows, decompress with PeaZip or the zstd binary from facebook/zstd releases. Checksums are in SHA256SUMS.

Quick start

-- A hadith's text. bhid = '{book_id}_{hadith_num}'; 146 is Sahih al-Bukhari.
SELECT clean_text FROM hadith_text WHERE bhid = '146_1';

-- Full-text search. The index is stored without tashkeel, so search without it too.
SELECT f.bhid, b.name
FROM hadith_fts f
JOIN hadith_text ht USING (bhid)
JOIN books b ON b.book_id = ht.book_id
WHERE hadith_fts MATCH '"إنما الأعمال بالنيات"'
LIMIT 20;

-- The isnad of a narration with each narrator's rank.
-- rawy_ids runs from the compiler back to the companion.
SELECT n.id, n.name, n.wasf_rotba
FROM sanad s, json_each('[' || s.rawy_ids || ']') j
JOIN narrators n ON n.id = j.value
WHERE s.bhid = '146_1'
ORDER BY j.key;

-- Every narration of the same hadith across all books.
SELECT h.bhid, b.name
FROM hadith h JOIN books b ON b.book_id = h.book_id
WHERE h.cluster_id = (SELECT cluster_id FROM hadith WHERE bhid = '146_1');

Tables

Table	Rows	What it holds
books	1,400	Book metadata: author, death date, edition, publisher, hadith_count
categories	18	Book categories
hadith_text	824,841	Every narration: clean_text (with tashkeel) and raw_text (with the program's inline markup)
hadith_fts	824,841	FTS5 index over clean_text, tashkeel stripped
hadith	463,733	Takhreej data per narration: cluster_id, companion (sahabi_id), grading (hokm)
clusters	20,744	A hadith across all its narrations: type, overall grading, number of companions and occurrences
book_clusters	610,426	Which clusters appear in which book
taraf	314,875	Opening phrases (atraf) per cluster
sanad	638,237	Isnads: rawy_ids is a comma-separated list of narrators.id, plus a grading
hokm_text	49,212	The written reasoning behind a sanad's grading (id = sanad.sanad_id)
narrators	49,845	Biographies: names, kunya, nisba, tabaqa, birth/death, rank (rotba, wasf_rotba), whether in al-Bukhari/Muslim
critics	1,015	The scholars of al-jarh wa al-ta'dil
critic_opinions	259,698	What each critic said about each narrator (alem_id → critics, rawy_id → narrators)
authors	487	Biographies of the books' authors
commentaries	22,644	Shuruh by bhid. commentary_book: 1 Fath al-Bari, 2 Sharh al-Nawawi ala Muslim, 3 Awn al-Ma'bud, 4 Tuhfat al-Ahwadhi
topic_nodes	55,455	Subject trees; leaves point to a cluster_id
quran_verses / verse_location	6,055 / 6,236	Quran text and surah/ayah lookup
dictionary	13,331	Entries by root from al-Qamus al-Muhit (mo7eet), al-Fa'iq (faaq), al-Sihah (sa7a7), al-Nihaya (nehay), Lisan al-Arab (lesan), Tahdhib al-Lugha (tahzeeb)
word_forms	630,841	Morphology: word form → root, prefix, pattern, suffix
manuscripts, references_bib, masdar		Manuscripts, bibliography, and source books

The full schema with comments is extractor/schema.sql.

Codes

Grading (hadith.hokm, sanad.hokm): 0 صحيح, 1 حسن, 2 ضعيف, 3 ضعيف جداً, 4 متهم, 5 موضوع, -1 not graded.

Narrator rank (narrators.rotba, label in wasf_rotba): 1 صحابي, 2 ثقة حافظ, 3 ثقة, 4 صدوق حسن الحديث, 5 صدوق يخطئ, 6 مقبول, 7 مجهول الحال, 8 ضعيف الحديث, 9 مجهول, 10 متروك الحديث, 11 متهم بالوضع, 12 وضاع.

Inline markup in raw_text: narrator names are wrapped as /94 name L<id> /94 (first narrator), /26 … /26 (chain) and /93 … /93 (companion); /20 starts the matn; /4 … /4 wraps a Quran quotation. render_hadith_html() in the site's api.php turns it into HTML.

License and attribution

This data comes from Jawami' al-Kalim, created by Offok and published freely by Islamweb as a waqf. The classical texts themselves are in the public domain; the gradings, takhreej, narrator assessments and indexes are their compilation. It is republished here in an open format so it can be used beyond the Windows app. Please credit Jawami' al-Kalim (Offok / Islamweb) in anything you build with it.

The license is marked "other" because the data isn't ours to license. Only the extraction code is MIT. If you represent the rights holders and have concerns, please open a discussion on this repository.