1 — INDEX
Point at Iceberg
The connector streams your Iceberg changelog into a derived full-text index — no second copy to keep in sync by hand.
2 — SEARCH
Query returns coordinates
Full-text, vector, and hybrid retrieval — plus filters, ranges, reranking, and even semi-structured Iceberg v3 variant columns. A search returns ranked coordinates — not a stale copy of your data.
3 — HYDRATE
Rows from the lake
Resolve coordinates back to the authoritative, governed rows in Iceberg by primary key. One source of truth.
Why not Elasticsearch or OpenSearch?
No second copy of your data. Your lakehouse stays the system of record; the index is derived and rebuildable — not a parallel datastore to reconcile.
The lake is the source of truth. Hydration returns the live, catalog-governed Iceberg row, not a search-time snapshot that drifts.
No reindex-the-world migrations. Point at a table and go; the connector keeps the index current from the Iceberg changelog.