A quick query takes a question in plain words, resolves the names to golden records through the cross-reference, applies your policy and generates SQL against the bed; the answer cites the rows it used and the query is yours to edit and save.
-- generated, governed, editable SELECT s.name AS plant, SUM(l.open_qty) AS open_qty, SUM(l.open_value_eur) AS open_value_eur FROM bed.sales_order_line l JOIN golden.customer_xref x ON l.customer_source_key = x.source_key JOIN golden.site s ON l.plant = s.sap_plant WHERE x.golden_id = 'L-00731' AND l.period BETWEEN '2026-Q3' AND '2026-Q3' GROUP BY 1 ORDER BY 3 DESC; -- resolved: Acme → L-00731 (4 source keys) · policy: row filter company code 1200 · 184 rows · 0.9 s
The generated SQL is shown, editable and saveable; a click on the resolution line shows the four source keys behind Acme; a click on the policy line shows the row filter that applied. The answer opens as a grid and can become a pivot or a chart with one click.
Through JDBC and ODBC the bed is a database: golden views per domain (golden.customer, golden.supplier, golden.material), the cross-reference tables, the hierarchies with their versions, the calendars, and the transactions read in place or from the governed replica. AS OF works on every table. The security engine rewrites the query with your policy before it runs.
Saved queries are objects: versioned, shareable, callable from the API as rows, usable as the source of a pivot, a chart or a rule.
A language model reads the question and the model of the bed (names, measures, dimensions); it sees no row data. The SQL it writes runs under your policy. See AI-assisted.
It is shown, so a reader can check and edit it; every answer cites its rows; a saved query is reviewed before it is shared.
ChatLima is the same quick query inside the conversational interface, with the same policy and citations.
Bring two systems that hold the same customers, suppliers or materials. In one working day we connect them, run the match on live rows, review the queue with your steward and publish the first golden view with its API.