SQL editor — Docs

Write SQL against every synced table and model, using DuckDB's dialect, and know the limits.

The SQL page is a full SQL editor over everything in your workspace. Your source systems' reporting limits don't apply: you can join across systems, use window functions, and query years of history.

The editor

Autocomplete covers every synced table, column and model in the workspace.
Run with ⌘↵ (Ctrl+Enter on Windows).
The Tables panel lists your tables, grouped by connection.
Explain describes what the selected SQL does, in a few sentences.
Save as model turns the query into a model.
Work is kept in named tabs that are saved on the server.

Dialect

Datachef runs DuckDB's SQL dialect. It's close to PostgreSQL, with useful extras:

-- Group by every non-aggregated column
SELECT region, plan, sum(mrr) FROM sample.customers GROUP BY ALL;

-- Filter on a window function without a subquery
SELECT * FROM salesforce.opportunity
QUALIFY row_number() OVER (PARTITION BY accountid ORDER BY closedate DESC) = 1;

-- Exclude a column from SELECT *
SELECT * EXCLUDE (description) FROM salesforce.account;

Naming tables and models

What	How to refer to it	Example
A synced table	`<connection name>.<table>`	`salesforce.opportunity`
A database table outside the default schema	`<connection name>.<schema>_<table>`	`warehouse.sales_orders`
A model	Its name	`active_customers`

See connection names.

Read-only

Every query is read-only:

One statement per query, and it must be a SELECT (including WITH ... SELECT).
Statements that write data or reach outside the workspace are refused. That includes ATTACH, COPY, INSTALL, PRAGMA, and functions that read files or URLs.
A query can only see tables and models in its own workspace.

Limits

Limit	Value
Rows shown in a result	10,000. A larger result is marked as truncated.
Rows in a CSV export	1,000,000
Query time	60 seconds

If you hit the row limit, aggregate in SQL instead. Summing or counting in the query is faster than pulling the rows out.