The Data Query Tester runs a data query on its own, outside the widget, sheet, or automation that will eventually hold it.

That separation is the point. A query that returns nothing inside a widget gives you no way to tell a broken query from a working query with no matching data. Here you see the result directly.

• Access
• The editor
  • Autocompletion
  • Keyboard shortcuts
• Results
• Agent pane

Access

Click Setup » Developers » Data Query Tester. This tool is limited to administrators – anyone else gets a permission error rather than a hidden menu item.

The editor

The query editor is Cerb UI's own, replacing the Ace editor used before 12.0.

It highlights two grammars at once, which is what a data query actually is: the query syntax itself, and any scripting tags embedded in it. Tag state carries across lines, so a multi-line {% ... %} block keeps its highlighting.

That isn't cosmetic. The tester evaluates the template before running the query, so a malformed tag fails before the query is ever executed. Nothing is in scope while it does – there are no placeholders here the way there are in a widget or an automation – so literals and pure functions work, and anything expecting a record won't.

Autocompletion

Suggestions come from four places, depending on where the caret is:

That last one is the biggest practical gain over the old editor. Inside query:(...) the fields descend exactly as they do in a worklist search bar – nested subqueries, deep filters like sender:org:, all of it – instead of leaving you to remember the grammar unaided.

Press Mod+Space, or use the Suggestions button on the editor's toolbar, to ask for suggestions anywhere.

Keyboard shortcuts

Mod is ⌘ on macOS and Ctrl elsewhere.

There's no shortcut for running the query – use the Run button.

Results

Results arrive as pretty-printed JSON in a read-only Cerb UI JSON editor: highlighted, foldable at any bracket with Mod+[ and Mod+], and searchable with Mod+F. Long lines scroll rather than wrap, so the structure stays readable.

The response is whatever the data query itself returned, with no wrapper around it.

Nothing here is cached. Every Run executes the query fresh, which is what you want while iterating – but it does mean the tester can disagree with a widget showing you a cached copy of the same query.

The tester imposes no limits of its own. What bounds a large result is the data query's own paging, such as limit:, and ultimately PHP's execution time and memory settings.

Agent pane

The Data Query Tester can host an agent pane – a chat that splits the screen beside the editor and can work on the query in front of it. Drag the divider to change the balance.

Enable an agent here by giving it a data_query block under components: in the AI tab of its worker record:

components:
  data_query:
    description: Help writing data queries

The agent's automation must use the interaction.worker.agent trigger, which is the only one offering the uiCommand element these commands run through.

An agent here has five commands:

editField matters more than it looks. An agent that can only replace the whole query has to reproduce every line it isn't changing, which is how a long query loses a filter nobody asked it to touch. A single-snippet edit changes what it names and leaves the rest alone.

Because this page never saves anything, an agent's edit exists only in the browser. Once one lands, leaving the page warns you first – which is the only thing standing between an agent-written query and a stray reload.