How this library is built
Source-grounded, traceable and honest about what is still unverified. Here is how records are made, connected and labelled.
The pipeline
Every page is kept
Each PDF page of a source is stored with its raw converted text and an automatic corrected reading. The printed table of contents becomes the topics of each session.
Curated extraction
Agents read bounded sections, four at a time, and write records for formulations, remedies, procedures, conditions, ingredients, principles and questions. Every claim keeps its chapter, page, verse, layer and the original wording.
Merged and connected
Records are merged by stable id, synonyms resolve to one identity, references are checked, and every relationship is stored in both directions. Unknown references are kept as marked stubs rather than dropped.
Indexed for discovery
Names from the lexicon and the records are matched on every page, so even chapters that are not yet extracted can be searched and explored. These matches are labelled as automatic.
One master file
Everything is exported as a single JSONL file with typed records, relationships, pages and a validation report, ready for a future database or search engine.
Nothing invented
No dosage, duration, efficacy, botanical identity or safety claim is added from outside the text. Missing information stays missing, and conflicting readings are both kept.
Labels you will see
- Traditional description
- What the classical text says, reported in neutral words. It is not a claim that the statement is scientifically established.
- Sanskrit verse
- The statement comes from the author's own Sanskrit verse.
- Hindi commentary
- The statement comes from the Bhāvārthadīpikā commentary, which sometimes explains or departs from the verse.
- Editorial note or footnote
- Remarks of the editor printed with the text.
- Needs a trained practitioner
- Purgation, emesis, enemas, bloodletting, cautery, caustics, surgery, metal and mineral preparations, toxic plants and poison treatment.
- Identity uncertain
- A plant name that could mean more than one species. No botanical identity is asserted.
- Reference entry
- A canonical name in the library that the curated extraction has not reached yet. Its page mentions are still listed.
- Automatic index
- Built from the structure of the book: contents, page text and name matches. Useful for finding, not for quoting.
Reading traditional knowledge safely
Ayurkosha is an educational library. It reports what traditional texts say, not medical advice. Many classical procedures need a trained practitioner, and some plants and preparations are unsafe. Talk to a qualified professional before using any remedy, especially during pregnancy, for children or alongside other medicines.
- Historical procedures are described for study, not as instructions to follow.
- Classical disease names are traditional categories and are never equated with modern diagnoses.
- Quantities are shown only as written in the source and are not dosage advice.
- Herb and medicine interactions, toxic plants, heavy metals and mineral preparations call for expert guidance.
- Symptoms that could be serious need prompt assessment by a qualified clinician.