With the infinite Kirpa of Sache Paatshaah, this project was undertaken as a humble Seva for the Panth. The goal of SikhLibrary is to provide a universal, comprehensive, and easily accessible digital source for Sikh research.

By centralizing scriptures, historical texts, scholarly commentaries, and linguistic archives into a standardized, machine-readable format, we aim to empower the next generation of Sikh scholars, AI researchers, and developers. If Sache Paatshaah wishes it, this dataset will serve as the foundational backend source for sikharchive.net—a global, living, centralized database dedicated to preserving open-source Khajana (treasure) of the Sikh Panth, making our profound literature, heritage, philosophy, and ethics transparently accessible to anyone in the world.

🚀 What's New in V3.0 (The Mahan Kosh Multilingual Expansion)

V3.0 brings the complete multilingual Mahan Kosh into the dataset — Bhai Kahan Singh Nabha's Gurushabad Ratnakar Mahan Kosh (1930), the great Sikh encyclopedia-dictionary, as structured JSON in 28 languages sharing one entry id across every language. Mirrored from the upstream repository redroyals/mahan-kosh-multilingual (Sikhi.io), released in honour of Guru Tegh Bahadur Ji's 350th year of sacrifice for the freedom of conscience.

Roughly 690 MB of new data. 64,218 live entries (68,150 rows including excluded front matter). Everything lands under Granths/Mahan_Kosh_Multilingual/.

`data/entries/` — one dictionary per language, keyed by entry id

File	Contents	Live entries
`core.json`	The spine: `id`, Gurmukhi headword (`hw`), Latin transliteration (`tr`), volume, page, `sources`, and an `excluded` flag for front matter / tables / abbreviation keys	64,218 live / 3,932 excluded
`en.json`	English definitions — Punjabi University Patiala edition (2006–2011), OCR'd	64,214
`hi.json`	Hindi definitions (from the English)	64,218
`gurmukhi.json`	Original Gurmukhi dictionary prose (not a translation). Partial by design — the remaining ~6.3% could not be joined to the book lemma without guessing	60,196 (93.7%)
`spanish.json` · `french.json` · `german.json` · `portuguese.json` · `italian.json` · `chinese.json` · `arabic.json` · `urdu.json` · `russian.json` · `bengali.json` · `gujarati.json` · `tamil.json` · `telugu.json` · `persian.json` · `japanese.json` · `shahmukhi.json` · `polish.json` · `dutch.json` · `swedish.json` · `romanian.json` · `czech.json` · `hungarian.json` · `danish.json` · `korean.json` · `malay.json`	25 machine translations of the English spine (`google/gemini` family), each locked per entry id. Korean and Malay were fidelity-audited (Sept 2026) and bad-graded rows excluded	~63,400–64,218 each
`corruption_flags.json`	Entries flagged for broken `See` references or script junk in the source English	680 flagged

`data/entries/tr/` — headword transliterations

arabic.json · bengali.json · gujarati.json · hi.json · japanese.json · persian.json · russian.json · tamil.json · telugu.json · urdu.json — the Gurmukhi headword rendered into each script (10 files).

`data/pages/` — raw per-page text, keyed `{vol}-{page}`

arabic.json · bengali.json · chinese.json · english.json · french.json · german.json · gujarati.json · italian.json · japanese.json · persian.json · portuguese.json · russian.json · spanish.json · tamil.json · telugu.json · urdu.json — the source page text behind the entries, in 16 languages.

`data/` root

File	Contents
`manifest.json`	Per-language counts, resolution percentages, and pipeline notes
`unresolved.json`	4,022 live ids that have a headword (and almost always English) but no recovered Gurmukhi body — listed honestly, not guessed at
`duplicate-siblings.json`	16,178 ids whose English source definition is byte-identical to another entry (the structural reason a translated entry can look duplicated)

Also included

README.md, QUALITY.md (upstream data-quality audit, dated 2026-08-14), LICENSE, and site/ — a small React / Vite search UI for the data.

Provenance & honesty

English and Hindi are the careful spine. The Gurmukhi body is the original 1930 prose. The other 25 languages are machine translation on the English spine — useful for access, not a scholarly citation. Do not quote a machine translation as Bhai Kahan Singh's own words. Shahmukhi is the weakest method (mostly English-MT, with some rows script-converted from the Gurmukhi body). Full detail in Granths/Mahan_Kosh_Multilingual/QUALITY.md.

Licence note for this folder

The upstream multilingual Mahan Kosh is released under CC BY 4.0 (the 1930 Gurmukhi original is public domain; the English follows the Punjabi University Patiala edition). Its bundled LICENSE file is kept alongside the data in Granths/Mahan_Kosh_Multilingual/. The SikhLibrary dataset as a whole remains CC BY-NC-ND 4.0.

🚀 What's New in V2.1 (The Multilingual Scripture & Reference Expansion)

V2.1 back-fills content that lived only in the sikhi.io / sikharchive.net production database and was not yet mirrored here. It adds roughly 1.2 GB of new text:

Multilingual scripture translations — Sri Dasam Granth (Sri Hazoor Sahib) in 12 languages, Sri Sarbloh Granth in 12 languages, and the Faridkot Wala Teeka in 7 languages (~4,300 pages each), plus a clean verse-by-verse English translation of the Dasam Granth. Non-English translations are machine-generated (google/gemini-2.5-flash-lite) and marked accordingly in the source database.
Full-text OCR for Sri Gur Pratap Suraj Granth (11,180 pages) and eight steeks/granths that previously appeared here only as a short chunked sample — Guru Granth Darpan (Prof. Sahib Singh), Panth Prakash, Sri Nanak Prakash, the Sarbloh Granth Steek, Shabdaarth Sri Guru Granth Sahib (Parts 1 & 2), Bani Bhagtan Ka Tika, and Gurbilas Baba Sahib Singh Bedi.
~45 additional books — colonial-era histories and ethnographies (Crooke, Hari Ram Gupta, Lepel Griffin, Sohan Lal Suri, and others), Nirmala / Sevapanthi steeks and janamsakhis, and historical Punjabi–English dictionaries.
Reference & discourse corpora — a combined Punjabi dictionary (87,636 entries), katha transcripts (1,179 recordings across multiple channels and languages), a katha-terms glossary (5,774), the Manglacharan literature corpus (543), Gurbani.org vichar posts (423), sakhi stories (515), AI verse explanations, and detailed per-Guru biographical pages (3,016). These are provided as .jsonl (one record per line) alongside per-item .txt.

Text added in this pass is a raw dump of the production pipeline's output; it has not been re-proofread and may contain OCR / machine-translation artefacts.

📁 Companion / Upstream Repositories

redroyals/mahan-kosh-multilingual (GitHub, by Sikhi.io) — the canonical, actively-maintained home of the multilingual Mahan Kosh edition that is now mirrored into this dataset at Granths/Mahan_Kosh_Multilingual/ (see What's New in V3.0 above). The repo also carries a live search viewer (sikhi.io/dictionary) and is where corrections, recovered Gurmukhi bodies, and real human translations should be contributed. The older English/Hindi-only Mahan Kosh text dump also remains here under Granths/Mahan_Kosh_Dictionary/.

🚀 What's New in V2.0 (The Grand Literary Expansion)

This update marks the largest single expansion to the repository by far, transforming it from a core scriptural database into a massive library of individual author archives.

To process this unprecedented volume of data, we built an advanced pipeline using OpenRouter to manage high-throughput Optical Character Recognition (OCR) parsing and intelligent text extraction, utilizing state-of-the-art models including Gemini 1.5 Flash-Lite and Gemma 4 31B. This allowed us to successfully convert thousands of heavily formatted, scanned historical texts into clean, perfectly structured UTF-8 text files and structured JSON text-pages.

📊 Dataset Statistics & Directory Structure

The repository has grown exponentially to capture individual works across over 50 prominent Sikh scholars, historians, theologians, and classical texts.

📁 Core Directory Structure

Folder / Path	Content Description
`Gurbani_Core`	Foundation scriptures including Sri Guru Granth Sahib (112 Languages), Dasam Granth, and Sarbloh Granth.
`Steeks`	Exegesis and word-by-word meanings (Teekas), including the ShabadOS database and classical commentaries.
`Granths`	Major historical works such as the Mahan Kosh (incl. the full 28-language `Mahan_Kosh_Multilingual` edition), Suraj Parkash, and Panth Prakash.
`[Author Names]` (New)	Dedicated directories for dozens of foundational authors and scholars (e.g., Bhai Vir Singh, Prof. Sahib Singh, Giani Dit Singh, Sant Seva Singh, Dr. Ganda Singh, and historical Granth authors).

🙏 Acknowledgements & Gratitude

This project is a culmination of decades of work by various organizations and individuals dedicated to digital preservation. We offer our deepest gratitude to:

Sache Paatshaah: For providing the strength, resources, and vision to perform this seva.
ShabadOS & BaniDB: For their incredible work in providing structured Gurbani data and indexing that serves as a cornerstone for modern Sikh apps.
Sikhi.IO: For their vast archive of digitized texts and translations that form the bulk of the research corpus.
Bhai Kahan Singh Nabha & Punjabi University, Patiala: For the Mahan Kosh and its English edition, the backbone of the Mahan_Kosh_Multilingual expansion.
Panthic Organizations & Digitizers: To all organizations, past and present, who have labored to scan, type, and verify these historical records.

🤝 Future Collaboration & Contributions

This is a living dataset. We believe that Sikh knowledge should be open and accessible to all.

Continuous Updates: This V3 update is just the beginning. We will continue to run processing pipelines to add more .txt and JSON data as new manuscripts are localized and verified.
Call for Collaboration: We invite other Panthic organizations, scholars, and developers to contribute raw scans or data to this repository. Whether it is a new translation, an unmapped historical teeka, or a linguistic corpus, your contribution directly fuels the preservation efforts at sikharchive.net.

If you have data you wish to contribute or if you find errors in the current texts, please open an Issue or a Discussion on this repository.

📜 Usage & License

This dataset is provided under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

Exception: the Granths/Mahan_Kosh_Multilingual/ folder is mirrored from an upstream project released under the more permissive CC BY 4.0; see the LICENSE file inside that folder.

What you may do

Download and share the files in their original, unmodified form.
Use the dataset for personal study, academic research, and non-commercial purposes.

What you may not do

Modify, alter, translate, or otherwise adapt any of the text files in this dataset.
Use this dataset or any portion of it for commercial purposes.
Redistribute any modified or derived version of these files.

Attribution requirement

Any use of this dataset must credit the following:

SikhLibrary Dataset (jsdosanj/SikhLibrary) on Hugging Face
Contributing sources: ShabadOS, BaniDB, Sikhi.IO
For the Mahan Kosh multilingual data: Bhai Kahan Singh Nabha, Gurushabad Ratnakar Mahan Kosh (1930); English after Punjabi University Patiala; packaged by Sikhi.io
Project maintainer: Bhai Jasvant Singh Panchhi (ਪੰਛੀ)

🛠️ How to Load (Python)

from datasets import load_dataset

# Load the entire archive
dataset = load_dataset("jsdosanj/SikhLibrary")

Downloads last month: 71

Total file size:

11.1 GB

You need to agree to share your contact information to access this dataset

🪯 SikhLibrary: A Universal Comprehensive Research Corpus

🛡️ Mission & Vision