What Sri Guru Granth Sahib Ji is

Explanation, not scripture — scholar review pending. These pages explain the scripture to engineers and students. Every quoted line is verbatim from the pinned database and cited by Ang; everything else is explanation, written by engineers and awaiting a Granthi’s review. Where this page and the scripture differ, the scripture is right.

Sri Guru Granth Sahib Ji is the scripture of the Sikhs and, since 1708, their living Guru: the Sikh tradition treats the text itself as the Guru, which is why it is always named in full, kept with reverence, and never edited. It holds the hymns of six of the ten Sikh Gurus together with those of saint-poets from several traditions — the Bhagats — and the praise-verses of the Bhatts, compiled by Guru Arjan Dev Ji in 1604 and completed by Guru Gobind Singh Ji. The hymns are in Gurmukhi script in a range of languages, and they are meant to be sung: most of the book is ordered by raag, the musical mode.

Every copy has the same 1,430 Angs — the word means “limb”, used instead of “page” — so a citation by Ang points to the same place in every printed volume. The project’s citation form is always the full name and the Ang: Sri Guru Granth Sahib Ji · Ang N.

The opening line

The Granth opens with the Mool Mantar, the statement of the One:

ੴ ਸਤਿ ਨਾਮੁ ਕਰਤਾ ਪੁਰਖੁ ਨਿਰਭਉ ਨਿਰਵੈਰੁ ਅਕਾਲ ਮੂਰਤਿ ਅਜੂਨੀ ਸੈਭੰ ਗੁਰ ਪ੍ਰਸਾਦਿ ॥
ik oankaar sat naam karataa purakh nirabhau niravair akaal moorat ajoonee saibhn gur prasaad
— Sri Guru Granth Sahib Ji · Ang 1

and Japji Sahib begins immediately after it, on the same Ang:

ਆਦਿ ਸਚੁ ਜੁਗਾਦਿ ਸਚੁ ॥
aad sach jugaad sach
— Sri Guru Granth Sahib Ji · Ang 1

The first line is exactly what /api/ang/1 returns as its first record — verbatim, with the Ang. The transliteration under each line is a reading aid built by the pipeline, not scripture (Transliteration and the fold).

This edition

The corpus reproduces one printed Bir — Siri Guru Granth Sahib in Gurmukhi with Index (user PDF, 1483 pp) — character for character: sggs-data’s reconcile.py proves the extracted text equals the PDF’s text stream, and every rebuild aborts on a single difference (Corpus pipeline and gates). Editions differ in small typographic conventions (the spacing of some words, the choice between the two nasal signs, the ligature used for the subjoined ya). The project does not “correct” its edition to another; it records which edition it is and notes differences where they are known.

The one rule

The text is never altered. Not corrected, normalised, paraphrased, reordered or guessed at — not in the corpus, the database, the API, the website, the app, search, copy or share. If a line looks wrong, it is flagged for a scholar and left as printed. The only transforms the pipeline applies decode the PDF font’s glyph order into logical Unicode, and every one of them is registered and reviewed in the editorial ledger.

Everything else in the system follows from that rule: the character-exact proof, the pinned dataset, the read-only database, the verification engine that never asserts a line it is unsure of, the app that refuses to show scripture if its bundled database does not hash to the manifest, and the wiki’s own gate that verifies every quoted line on these pages against the database.

Two questions

Read next: The structure.