📜 isnady  _إسناد

Hadith search built around the isnad, the chain of transmission
_{Search the Sunni and Shia collections, read every chain narrator by narrator, and see how the scholars of each tradition read it}

Repositories • Why • Install • Quick Start • isnady-data • Search • Chains • Sunni and Shia • Narrators • Books • Scholars • Statistics • Learn • Meaning (AI) • Appearance • CLI • Roadmap

Pre-alpha (0.2.0). Search, chains of transmission, narrators, books, scholars, statistics and the glossary work today, Sunni and Shia, on data you import in one click. Everything is built in the open, one piece at a time.

🧭 Two repositories, one project

isnady is two repositories that work as one:

   bayramkotan/isnady-data                         bayramkotan/isnady
   ───────────────────────                         ──────────────────
   catalog.json ─────────────── Data sources ───▶  import ─▶ isnady.db ─▶ search · chains · narrators
   sunni/<book>/  shia/<book>/                       (your data folder)      books · scholars · statistics
   rijal/sunni/   rijal/shia/                                                 window  and  command line
   Releases: isnady-db-<date>.zip ── Use a ready database… ──▶ isnady.db

The application ships without hadith data, so it stays small and the data can grow on its own: a new book in isnady-data reaches every isnady at once, without a new version of the program.

🎯 Why isnady

A hadith is two things: the text (matn) and the chain of people who passed it on (isnad). Most hadith sites let you search the text and show a grade. isnady is built the other way round: the chain comes first, because that is where the scholars of hadith did their work.

• The chain is read, not typed in. isnady reads the chain from the Arabic text itself: who narrated to whom, with which words, and whom it reaches.
• Each tradition by its own road. A Sunni book's chains are read the Sunni way and judged by the Sunni critics; a Shia book's chains are read the Shia way and judged by the Imami critics. The two meet on the same pages and are never measured by each other's scale.
• Nothing is guessed. Where the wording cannot be read with confidence, the chain is kept whole and marked, never split wrongly. A narrator is named only when the evidence leaves one person; otherwise isnady says how many it could be.
• Every fact has a source. Every hadith, text and grade records where it came from, under which licence.
• Grades are shown as the scholars gave them. Each grade keeps its scholar's name and wording; grades are ordered categories, never averaged, and isnady never invents its own.
• Computed once, kept. Every statistic and model is saved as JSON beside your data and read back, never computed again for the same data.
• Window and command line, same engine. Everything the window does, the command line does too, with the same results.

📦 Install

Desktop application

Download isnady for your system from the latest release — everything included, the AI features too:

All three are portable: nothing is installed, each runs from wherever it is. Put a folder named isnady-data beside the file and isnady keeps its databases and settings there — isnady on a USB stick.

Put isnady on the desktop and in the applications menu with Tools → Create Desktop Shortcut (or iy shortcut): the shortcut starts isnady the way it is installed — the AppImage, the application, or the isnady-gui command of a Python install.

With Python

One line, on any system. Install isnady wherever you like — for all users, for yourself, in a virtual environment, with pipx: the same line installs it, and later updates every copy where it is, asking for administrator rights only when a copy needs them. Nothing is ever removed.

# Linux and macOS
curl -fsSL https://raw.githubusercontent.com/bayramkotan/isnady/main/install.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/bayramkotan/isnady/main/install.ps1 | iex

Or with pip:

pip install isnady
isnady            # or the short name:  iy

Run inside a clone of this repository, the installer makes an editable (developer) install of that clone instead.

The same program answers to several names: iy to type, isnady to read, and isnady-cli kept from the first releases. Without arguments it opens the window; with arguments it is the command line. On Windows, isnady-gui opens the window without a console.

# Isolated — recommended, no system packages touched
pipx install isnady

# Into your user site — needs the override flag
pip install isnady --break-system-packages --no-cache-dir -U

Upgrading, and when something is not right

iy update

iy update updates every copy of isnady where it is installed. Your data stays where it is; a newer isnady upgrades the database in place the first time it opens it, and reads the chains again when its chain reader has changed (a minute or two, once).

iy doctor lists every copy and which one each command really starts — an older copy in the user folder can start before a newer one elsewhere, so every copy is kept at the same version. The same report is under Help → Check Installation. If an old copy starts even for iy update, run python3 -s -m isnady update (Windows: py -s -m isnady update).

🚀 Quick Start

1. Open isnady — iy, or the desktop application.
2. File → Data Sources. The repository is already chosen: isnady-data. Tick the books and the languages you want and press Import — or Import all.
3. Import the narrators too — Ibn Hajar's Taqrib for the Sunni books, al-Najashi's Rijal for the Shia books — and every name in every chain is matched to its narrator.
4. Search, open a chain, open a narrator, read a book.

The same from the command line:

iy catalog list                                           # the books of isnady-data
iy catalog import sahih_al_bukhari --language ara --language tur
iy catalog import musnad_ahmad_ibn_hanbal
iy catalog import al_kafi_al_kulayni                      # with al-Majlisi's and al-Behbudi's grades
iy catalog import taqrib_al_tahdhib_ibn_hajar_al_asqalani # the narrators of the Sunni chains
iy catalog import rijal_al_najashi_ahmad_ibn_ali_al_najashi  # the narrators of the Shia chains
iy catalog import --all                                   # everything

iy search النيات              # Arabic, with or without diacritics
iy search "niyetlere göre"     # Turkish, English, any imported language
iy chain bukhari 1             # one chain, narrator by narrator
iy chain kafi 5:1740           # a Shia chain, read the Shia way
iy narrator الزهري             # who a narrator is

Importing takes a minute or two a book; the search index, the chains and the narrators are built as part of it. No network? See Without the network below.

📚 isnady-data — the data

bayramkotan/isnady-data is isnady's own data repository and its default: every source isnady reads, converted once into isnady's own format.

How it is laid out

isnady-data/
├── catalog.json                    the list isnady reads (isnady-catalog/1)
├── sunni/sahih_al_bukhari/         book.json (numbers, chapters, references, grades) + ara.json, eng.json, tur.json …
├── sunni/musnad_ahmad_ibn_hanbal/
├── shia/al_kafi_al_kulayni/
├── rijal/sunni/taqrib_al_tahdhib_ibn_hajar_al_asqalani.md
├── rijal/shia/rijal_al_najashi_ahmad_ibn_ali_al_najashi.md
├── derived/                        statistics and models already computed, as JSON
└── database.json                   the ready database of the latest Release

• Full names, never a bare one. Every folder and every id is the book's full title with its compiler's name — musnad_ahmad_ibn_hanbal, al_kafi_al_kulayni, taqrib_al_tahdhib_ibn_hajar_al_asqalani. isnady imports by these ids, and names its exports and saved computations after them.
• One format for every book (isnady-book/1): a book.json with each hadith's number, place, chapter, other numberings and every grade as [scholar, grade], exactly as written; and one file per language with the chapter titles and the texts. The formats are described in the isnady-data README.
• Sunni and Shia apart. Every book says its tradition, and isnady keeps them on their own roads (below).
• Your own data stays yours. What you import from your own files is never written into a ready database for isnady-data: it does not leave your computer.

Without the network

isnady-data works the same from a folder on your computer — a git clone, a USB stick, a shared folder:

git clone https://github.com/bayramkotan/isnady-data ~/isnady-data
iy repo add ~/isnady-data       # in the window: Data sources → A folder on this computer…
iy catalog import --all

Or take a whole ready database — every book imported, every chain read, every narrator identified — from the isnady-data Releases:

iy database use ~/Downloads/isnady-db-<date>.zip   # in the window: Data folder → Use a ready database…

isnady checks the file (the zip must hold isnady.db, the schema must be one it reads, the sha256 must match the database.json beside it), puts it in place at the next start and keeps your previous database beside it (isnady.<date-time>.bak.db). Nothing is deleted.

Other repositories

isnady-data is all you need. isnady can also read a few datasets that others publish, as they publish them, under their own keys (e.g. bukhari-lk) so that editions with other numbering never mix with isnady-data's. Choose one in Settings → Data sources (iy repo use ID) or add any repository with an isnady catalog, a Hugging Face dataset, a URL or a folder (iy repo add …):

Every source records a licence tier — A may be redistributed, B under its conditions, C may not — and Help → Licences lists them all. Any CSV or JSON table of hadith can be read through a catalog that names its columns (format table); files and links of your own, with or without authentication (credentials are never stored), are read as well.

🔎 Search

Results appear at once and fill in as you read; the window never waits for them.

🔗 Chains of transmission

Every Arabic text is read for its chain:

Sahih al-Bukhari 1
   1. حدثنا (narrated to us)  الْحُمَيْدِيُّ عَبْدُ اللَّهِ بْنُ الزُّبَيْرِ
   2. حدثنا (narrated to us)  سُفْيَانُ
   3. حدثنا (narrated to us)  يَحْيَى بْنُ سَعِيدٍ الْأَنْصَارِيُّ
   4. أخبرني (informed me)    مُحَمَّدُ بْنُ إِبْرَاهِيمَ التَّيْمِيُّ
   5. سمع (heard)            عَلْقَمَةَ بْنَ وَقَّاصٍ اللَّيْثِيَّ
   6. سمعت (I heard)          عُمَرَ بْنَ الْخَطَّابِ
      -> the Prophet ﷺ

In the window, each result shows its chain as a row of names ending at whom it reaches, and View chain opens it as a timeline from the book to the Prophet ﷺ — or, in a Shia book, to the Imam — with the Arabic text below: the chain in lighter ink, the text of the hadith in full ink. Every link says who the narrator is and what his critic says of him; a click opens him.

👤 Who each narrator is

Every name in a chain is matched to a narrator of its own tradition's rijal work — the Taqrib for a Sunni book, al-Najashi for a Shia book. A name is linked only when the evidence leaves one person: the words of the name, the book marks (a narrator in al-Bukhari must be one Ibn Hajar marks خ), the order of generations along the chain, the last link before the Prophet being a Companion, and — in al-Kafi — al-Kulayni's own teachers at the head of his chains. When several narrators remain — "Sufyan" can be al-Thawri or Ibn 'Uyayna — the chain says so and names how many, rather than choosing one. In blind, hand-checked samples of the identified names, the last fifty were all correct. Names are kept exactly as written, with the clarifications the compilers added ("— yaʿnī Ibn Muḥammad —").

🎓 The words of transmission explain themselves

🕌 Sunni and Shia: two roads

Sunni and Shia meet on the same pages — one search, one Narrators page, one shelf of books — and every Shia narrator and book carries a gold Shia tag. They meet without being mixed: every value isnady derives is reached along the road of the book's own tradition.

Nothing is mapped onto the other tradition's measure, and a narrator of one tradition is never a candidate for a link of the other's chains. Read this way, 96.8% of al-Kafi's 14,239 chains split into their narrators (9 did when they were read the Sunni way), 92% reach a Ma'sum, and half of their 61,160 names are identified in al-Najashi's Rijal (3 were before).

iy chain kafi 5:1740                                 # al-Kulayni's group, al-Najashi's verdicts, the Imam it reaches
iy search "زرارة" --in narrators --tradition shia
iy isnads                                            # per book: chains split, and whom they reach

👥 Narrators

The Narrators page lists every narrator of the imported rijal works — All, Sunni or Shia at the top of the list — searchable in Arabic or in Latin letters (Abu Hurayra, Zuhri, Ibn Umar, زرارة). For each narrator:

• the name he is known by (al-A'mash, Ibn 'Umar, Zurara), with his full name and every part of it laid out — name, lineage, kunya, by-name, nisbas, client of — the part he is known by marked ★;
• what his critic says, in the critic's own words: Ibn Hajar's verdict and its rank for a Sunni narrator; al-Najashi's words, reliability and creed for a Shia narrator — the entry as written one click away;
• tabaqa and death year, the books his hadith appear in (Ibn Hajar's marks);
• whom he narrates from and who narrates from him in the imported chains, and the compilers who narrate from him directly;
• every hadith whose chain includes him, each a click from its chain.

Sunni narrators filter by tabaqa, rank and book; Shia narrators by the Imami assessment.

📖 Books

isnady is not only for searching: the Books section opens every imported work and reads it in its own order. A hadith collection reads chapter by chapter, each hadith with its card (chain, narrators, other narrations, grades) and the languages you choose; Ibn Hajar's Taqrib and al-Najashi's Rijal read entry by entry, each narrator a click away. Find within a chapter, go to the previous or next one, and isnady remembers where you were. From any search result, In its book opens the hadith where it stands, among the hadith of its chapter.

The shelf holds the Sunni and the Shia books together, each under its own heading — All, Sunni or Shia at the top shows one of them alone — and every Shia book carries a gold Shia tag. Books are not tied to hadith collections: a work is a tree of any depth (volume, book, chapter, section) whose leaves are hadith, narrators' entries or paragraphs — so commentaries, manuals of fiqh and the other works of the scholars can be read the same way as they are added.

iy book                      # the books
iy book bukhari              # its chapters
iy book bukhari 2 --language Turkish
iy book taqrib 1

🎓 Hadith Scholars

The scholars whose work is in the imported data, and what the data measures of it: compilers (their collection, the teachers their chains begin with, the Companions they end with), graders (how they grade, and how far they agree with each other on the same hadith — agreement, Cohen's kappa, and a strictness index: the classical mutashaddid / mutasahil, measured), and critics of narrators — Sunni and Shia on one page (All / Sunni / Shia), each by his own road: al-Kulayni, al-Saduq, al-Mufid and al-Tusi with their chains counted to a Ma'sum; al-Majlisi and al-Behbudi compared only with each other on the four Imami grades (sahih, hasan, muwaththaq, da'if); al-Najashi's verdicts in the Imami terms. On Sunan Abi Dawud, for example, Zubair 'Ali Za'i grades most strictly (−0.21) and agrees with al-Albani on 70% of the hadith; the page also flags a pair of graders whose 98% agreement suggests the source did not keep them independent.

📊 Statistics

Measured in depth, each figure with what it means in plain words and how sure it is — on two roads: All Sunni books (Ibn Hajar's ranks, chains to the Prophet ﷺ) and All Shia books (al-Najashi's Imami categories, chains to a Ma'sum; al-Majlisi's and al-Behbudi's grades in the four Imami categories — sahih, hasan, muwaththaq, da'if), never mixed. Every section is measured on both roads. Grades are ordered categories, so they are never turned into numbers or averaged, and no figure is read against the science of hadith: the statistics measure what the scholars did, they do not overrule it. Every result is computed once and saved as JSON in your data folder (derived/).

iy stats                     # what the database holds, per book
iy stats graders --book abudawud
iy stats model --grader albani
iy stats lab --csv ./lab     # every table behind the figures, as CSV

🎓 Learn

The terms of hadith and its sciences, written for isnady with their classical sources: 81 terms — the grades (sahih, hasan, da'if, mawdu', shadhdh, munkar, mu'allal), the kinds of hadith by their chain (mutawatir, mursal, munqati', mu'allaq, mudallas …), chain and text (mutaba'a, shahid, takhrij, madar), the ways of receiving (sama', ijaza, haddathana, 'an), judging narrators (the twelve ranks of Ibn Hajar, thiqa, saduq, majhul, matruk, mudallis), generations, the parts of an Arabic name, the Imami grades (sahih, hasan, muwaththaq, da'if), the books, and isnady's own measures. Each with its Arabic, a one-line and a longer definition in English, Turkish or Arabic — in the interface language when the glossary has it — related terms one click away, and its source. The grades on every result explain themselves when you point at them. iy term mursal --lang tr, iy term mursal --lang ar.

🧠 Search by meaning (AI)

Choose Match → By meaning (AI) to find hadith that say the same thing in other words or in another language: komşu hakları finds the hadith on the neighbour's rights, a Turkish sentence finds its Arabic original, and إنما الأعمال بالنيات gathers the narrations of the hadith on intentions from every imported book. Each result shows how close it is in meaning (0–1), which says nothing about authenticity.

The model is learnt on your computer from the texts you imported — Arabic beside its translations is a parallel corpus, and cross-lingual latent semantic analysis learns the concepts the languages share. Nothing is downloaded and no outside model is used. Measured on al-Bukhari and Abu Dawud: for hadith the model never saw while learning, a Turkish translation finds its Arabic original first 68% of the time and among the first ten 93% of the time (chance: 0.8%). For an exact quotation the ordinary word search stays the better tool.

The same step finds the other narrations of each hadith across the imported books (takhrij): every result card says where else it is narrated — Also narrated in: Sahih al-Bukhari 5070 · 6689 · 6953 | Sunan Abi Dawud 2201 — and each number opens that hadith. Only the texts are compared, never the chains. In blind, hand-checked samples 34 of 35 pairs of the first kind and 18 of 20 of the second were right.

pip install "isnady[ai]"        # numpy and scipy
iy ai build                     # a few seconds; or Tools → Build AI Indexes
iy search --mode meaning "komşu hakları"
iy tahric bukhari 1             # the other narrations of a hadith

🎨 Appearance and settings

Seven themes, chosen in Edit → Preferences → Themes — each shown as a small live isnady window — or in View → Theme: Lapis (lapis and gold of illuminated manuscripts) and Lapis Night, which "Follow system" switches between; Paper (warm cream and ink, for long reading); Slate; Emerald; Midnight; and High Contrast (black, white and yellow, for low vision).

Every script has its own reading font, size, colour and line spacing — Arabic, Latin, Cyrillic, Bengali and Tamil today. The light and the dark theme each keep their own colours, and the interface font can be changed too. Edit → Preferences (Ctrl + ,) applies every change at once; each row has its own Default button.

Interface language (sidebar, or View → Interface Language): menus, pages, buttons, messages and the statistics are in English, Türkçe and العربية — Arabic right to left, the whole window mirrored. The list already names 24 languages, each in its own name with how much of it is translated (Deutsch, Français, Español, Português, Русский, Bahasa Indonesia, Bahasa Melayu, हिन्दी, اردو, বাংলা, 中文, 日本語, 한국어, فارسی, دری, پښتو, Azərbaycanca, Oʻzbekcha, Bosanski, Kiswahili, Hausa); a text not translated yet is shown in English, and the translations follow release by release. The window is rebuilt in the new language at once, nothing to restart. iy config set view.ui_language tr. The hadith texts keep their own languages (Content languages), the command line stays in English.

Settings → Books and languages switches imported books on and off (Sunni and Shia apart) and chooses the languages the texts are shown in; the others are folded behind an arrow on each hadith.

iy config list --prefix text.arabic       # what is set, and what is changed
iy config set text.arabic.size 22
iy config set view.theme paper
iy config reset text.arabic               # back to the defaults

Your data folder

Everything isnady keeps is in one folder — iy datadir shows it, iy datadir /elsewhere moves it, and a folder named isnady-data beside a portable application is used instead:

⌨️ CLI

Everything the window does also works without it — on a server, over SSH, or in a script. The command line never loads Qt. Every command that takes a book accepts its full name (musnad_ahmad_ibn_hanbal) or its short key (ahmad).

$ iy search niyet --book abudawud --limit 1
43 hadith found in 8 ms

Sunan Abu Dawud #472
  grades: Al-Albani: Hasan; Muhammad Muhyi Al-Din Abdul Hamid: Hasan; Shuaib Al Arnaut: Daif; Zubair Ali Zai: Daif
  [Turkish] Ebu Hureyre (r.a.); Resulullah (Sallallahu aleyhi ve Sellem)'in şöyle buyurduğunu rivayet etmiştir: "Bir kimse mescid'e hangi [niyet]le gelirse nasibi ondan ibarettir"

📸 More screenshots

🗺️ Roadmap

• Shia narrators, further — al-Tusi's Rijal and Fihrist and al-Kashshi's reports; names resolved by teacher and student; a Sunni–Shia comparison.
• More in isnady-data — al-Tusi's Tahdhib al-Ahkam and al-Istibsar, al-Saduq's Man la yahduruhu al-faqih; al-Hakim's Mustadrak, al-Bayhaqi's Sunan, the Mu'jams.
• Narrators — the other half of the names identified through teachers and students (al-Mizzi's Tahdhib al-Kamal), the verdicts of other critics beside Ibn Hajar's, and whether each link could have met the next; fuller, richer lives of narrators and scholars.
• Statistics in plain words — every finding said simply first, the method after.
• Settings and export — every setting in one place, export in every common format (named by the books' full names), reset and remove from inside the window.
• Every interface language in full — the 21 listed after English, Türkçe and العربية; the command line and the glossary in them too.
• isnady.net — the same engine on the web.

🔁 Versions and releases

One version number on GitHub, PyPI and npm, always. A release builds the Python package and the Windows, Linux and macOS applications first; only when every build succeeds is anything published — PyPI, then the GitHub Release with every file and its checksum. What each version brought is in the CHANGELOG. The README on npm is made from this one. The data has its own rhythm: a book added to isnady-data needs no new version of isnady.

🏗️ Build from Source

git clone https://github.com/bayramkotan/isnady.git
cd isnady
pip install -e .
iy

Python 3.10 or newer and PySide6. The command line alone needs no PySide6. The desktop application of your own system: pip install ".[ai]" pyinstaller, then python packaging/build_app.py linux|windows|macos VERSION.

The data repository is built by two tools in this one: tools/build_data_repo.py --mirror DIR converts every source into isnady-data, and tools/publish_database.py DIR prepares a ready database for its Releases.

📝 License

MIT for the application code. Hadith data comes from separate sources, each under its own licence, recorded with the data and listed under Help → Licences.

The Arabic and reading typeface is Amiri by Khaled Hosny, bundled under the SIL Open Font License 1.1 (src/isnady/assets/fonts/OFL.txt).