title: What GRETIL is, and how we use it summary: The Göttingen Register of Electronic Texts in Indian Languages supplies the Sanskrit of most of the Library. What it is, what it is not, and the terms we hold it under. date: 2026-09-08
The archive
GRETIL — the Göttingen Register of Electronic Texts in Indian Languages — is maintained by the Niedersächsische Staats- und Universitätsbibliothek Göttingen. For decades scholars have contributed machine-readable transcriptions of printed editions to it: Vedic saṃhitās and brāhmaṇas, the Upaniṣads, the critical editions of the Mahābhārata and Rāmāyaṇa, Purāṇas, kāvya, and śāstra across grammar, law, medicine, astronomy and philosophy, together with a large body of Buddhist and Jaina literature.
Each file names the printed edition it was typed from and the person who input it. That provenance is carried into the Library and shown on every text's page.
What we import, and what we refuse
- Only units that carry a reference in the source file are imported. A verse we cannot cite is a verse we cannot verify, so unreferenced running text is left out.
- Pāda indexes, duplicate encodings (accented and unaccented Vedic text, "plain" and "with pada markers" versions) are skipped so nothing is counted twice.
- Where a text's numbering is inferred rather than stated in the file — a few texts numbered only within chapters — its provenance says so.
- Devanagari is generated mechanically from GRETIL's IAST; the IAST is the stored source of truth.
- About 470 GRETIL files remain unimported because their reference formats need their own parsers. The list is public in the import report.
Licence
GRETIL's TEI corpus is released under Creative Commons Attribution–NonCommercial–ShareAlike 4.0. HREEM holds and displays it with attribution, non-commercially, and keeps the Companion and the Library free. Anyone who holds rights in an e-text and objects to its presence can write to us and it will be removed.
What GRETIL is not
It is not translation. Nearly everything in it is Sanskrit only. English in the Library comes from separate public-domain sources — the open gita/gita dataset for the Gītā, Swami Paramananda (1919) and F. Max Müller (1879–1884) for the Upaniṣads, Georg Bühler (1886) for the Manu-smṛti — each aligned verse by verse under its own rule.
A parsing lesson worth recording
For weeks the Library held only 1,525 of Manu's 2,685 verses and nobody noticed, because the count looked plausible. Two small things in GRETIL's file were responsible: an alternative verse numbering in square brackets before the slash (Manu3.67a[57ṃa]/), and a single editorial note containing a bare "<" (mety < ma ity ?), which the HTML tag-stripper read as the start of a tag and swallowed everything to the end of the file. Both are fixed; the fix for the second protects every other GRETIL file too. The general point stands: a corpus is only as complete as its import report says, so we publish the report.