ponder(), embed()) run on the same connection.
How a tap authenticates and pages the source is on
the connector and
ingestion. The lake is where queryable tables
live.
Quality is a separate catalog,
providers_quality, not a fifth data pool. It holds SPC series, Gate 2
checks, table stats, and health for provider tables.
AI is ponder() and embed(): LLM and embedding calls in
the same SQL.
References are fully qualified. There is no USE or search_path. The
workspace on the request selects which tables you see; the catalog name
never encodes the workspace.
The catalog
The catalog lists every table across the four pools. Per table it reports:- Shape: row counts and schema.
- Freshness: when data last arrived (providers), publish time (artifacts), last successful Plot run (parables). Workspace tables are current at query time.
- Quality: automated checks and SPC. Query them in Quality.
- Lineage: which taps or Plots feed the table.
SELECT COUNT(*) FROM ... LIMIT 1 if you need to know whether data
exists.
Querying the lake
Lake tables are queryable over Arrow Flight SQL. JDBC and ADBC drivers work, as do DataGrip, DBeaver, and the clients below. You authenticate with the same API token you use for the REST API. This example queries the providers pool. Swap the SQL forartifacts.{table}, parables.{parable}.{plot},
workspace.{table}, or providers_quality.summary.* the same way.
Flight SQL access is provisioned per workspace. Contact your Parable
representative to enable it and receive your workspace’s query endpoint.