Occurrence records, museum collections, and the peer-reviewed literature answered as one corpus. The snapshot it answered from is versioned and dated, so a result can be reproduced later.

No written case study for this one yet. The Technical and Source readings carry what is on record.

Architecture

A mirrored, versioned snapshot of the source records; a warehouse of repartitioned parquet answered by a typed query DSL; a vector corpus of registry, taxonomy and literature documents; an agent loop that must cite what it returns or fail closed.

Built out of

Next.js
Corpus-grounded retrieval
Versioned snapshots

Evidence produced

Every answer names the snapshot date it was drawn from.

Known limitations

Common-name search is thin where the source vocabulary lacks vernacular names.

Status

Live. Deployed and answering right now. Started 2026.

Artifacts

LiveOpen the live app ↗

Running at

Open the live app ↗

Shares with

TRINIDAD →
Corpus-grounded retrieval with citation
DRAMATURG →
Entity resolution over records
Celaya Solutions Research →
Provenance-aware retrieval

Can occurrence records, museum collections and the literature be answered as one corpus, reproducibly?

ARK