Scripture Lab reads verse text and book metadata from PostgreSQL Bible databases. Every loaded translation is a public-domain text. Optional AI commentary is composed only from the verses retrieved for your question — never from the model’s memory alone.
How Ask answers are produced
- Your question is matched against verses in the databases using keywords, topics, and explicit references.
- The returned answer, explanation, and cited verses all come from that retrieved evidence pack.
- When an OpenAI key is set, commentary is generated from the pack; a citation checker strips invented Book chapter:verse references.
- Without a key, or when a prior verified answer exists, the app serves local or cached answers built from the same database evidence.
Where Scripture and books come from
- Full-text Bibles are ingested from public-domain sources — midvash bible-data, eBible USFM, JPS 1917, Brenton LXX, and the Ethiopian 81-book Ge'ez corpus — into versioned JSON and PostgreSQL schemas.
- The catalog lists tradition, language, and license for every referenced translation; copyrighted modern versions stay catalog-only and are not loaded as study text.
- Whole-book introductions and chapter guides on Books are curated study notes kept with the site, not generated on the fly. Every claim on Canon must cite a source.
How content is verified
- Verse text is always read from the database at request time — not paraphrased from model training data.
- Shared Ask answers are stored only after citation verification against the evidence pack.
- Chapter guides are checked for structure, uniqueness, and agreement with the Bible text before they are published.
- Corpus builds and migrations are versioned so operators can validate releases before publication.