Skip to Content
AIGeoEmbeddingsGeoParquet

Connect to GeoParquet

Download GeoParquet (and optional metadata JSON) that a GeoEmbeddings worker job uploaded to project Storage. This is for OLAP / DuckDB / pandas — not the online RPC API.

Artifacts exist only if the job set uploadArtifactsToStorage: true. Studio stores bucket/object paths under embeddings_catalogue.data_source.artifacts and mints time-limited signed HTTPS URLs on the table details panel.

Studio flow

  1. Open the embeddings table → Table details.
  2. If artifacts exist, use the signed links (refresh when expired).
  3. Connect to GeoParquet offers copy-paste snippets: curl, Python, DuckDB.

Typical usage

curl — download to disk, then open locally.

Python — pandas.read_parquet(url) or geopandas.read_parquet(url) with the signed URL (needs a working pyarrow HTTPS setup; otherwise curl first).

DuckDB — load httpfs + spatial, then read_parquet('…signed URL…').

Mint your own URL — with service-role credentials:

# conceptual; use supabase-py against GEOBASE_PROJECT_URL
supabase.storage.from_(bucket).create_signed_url(object_path, expires_in_seconds)
  • Create jobs: Worker jobs · createGeoEmbeddings / createSraiEmbeddings
  • Online search: RPC API
Last updated on
RPC APISRAI