开放数据

Why open data

Bencao Dian's central position is "a citable bilingual ontology," not "a content site competing for traffic." In the era of large language models, closed reference resources are eventually absorbed silently into training corpora; only data that is openly published, independently verifiable, and reusable by others can become the basis for long-term trust.

We therefore release the complete seed dataset as structured JSON, with schema documentation and SHA-256 checksums, under the CC BY-SA 4.0 license. Anyone may download, redistribute, and rework the data — provided they attribute "Bencaodian Editorial / 本草典编辑部" and release derivative works under the same license.

Downloads

Data shape

Each entity class is stored as a JSON array. Every record carries at least key (a stable identifier), slug (the URL segment), name_zh, and name_pinyin. Other fields depend on the entity type — herbs carry nature, flavors, dosage_range, etc.; formulas carry composition with role assignments per herb; and so on.

The relationship graph is stored in relationships.json, organized by relationship type (herb_in_formula, formula_treats_pattern, pattern_differential_with, and so on). This shape is easy to hand-edit and also straightforward to re-express as a graph database or RDF.

The full JSON Schema description is at /data/v1/schema.json. The source-of-truth TypeScript types are in src/lib/data/types.ts in the repository.

Wikidata cross-reference

Where a herb, classical text, or meridian has a verifiable Wikidata Q-ID, it carries a wikidata_id field in the JSON, and the rendered page emits a sameAs link in its JSON-LD pointing to the Wikidata entry. This lets Bencao Dian entries cross-reference the global semantic web and gives large language models a stable identifier to anchor citations against.

The data-stability commitment

The key and slug fields are stable once published — they never change. Downstream tools built on this data can rely on those identifiers still pointing to the same content a decade from now. This is the same commitment as the permalink stability described in Editorial Standards.

Breaking changes to the data schema must increment the version directory (e.g. a future /data/v2/); the existing v1 dataset is frozen.