Exploration
Inspect schemas and run read-only SQL to understand your data and draft analytical logic.
Exploration is where you understand your data and draft the logic a contract will later make stable. It's a builder-only mode, and it's kept separate from consumption on purpose.
What exploration is for
Before you can write a contract worth trusting, you need to know what the data actually contains: which columns exist, how they're typed, how clean they are, and what a correct calculation looks like. Exploration is where you answer those questions.
Exploration is iterative. You're meant to run lots of queries, get things wrong, and refine, because nothing here is live. The output isn't a result you ship. It's the understanding that lets you author a good model or contract.
Inspecting schemas
Open any table or model to see its schema: column names, data types, and metadata such as row counts. Schema inspection is the fastest way to learn the shape of a dataset before you query it.
Running SQL
The workbench lets you run read-only SQL against your tables and models.
Exploration SQL is SELECT-only. Statements that would modify data or schema
aren't permitted. Exploration observes data, it never changes it.
Use SQL exploration to:
- Profile columns: distinct values, ranges, null rates.
- Check how tables relate before you join them in a contract.
- Prototype the exact aggregation a measure should perform.
- Verify that a model's logic produces what you expect.
Results are capped at a sensible row limit so exploratory queries stay fast.
Asking questions in natural language
Exploration also has a chat interface. You can reference tables, models, and collections and ask questions in plain language; Gaur generates and runs the SQL for you and explains the result. It's handy for quickly profiling unfamiliar data without hand-writing every query.
From exploration to a contract
Exploration is a drafting space. The logic you discover here isn't governed yet. To make it stable and consumable:
Capture reusable logic as a model
If your logic involves cleaning, reshaping, or pre-aggregating data, save it as a model so it can be reused and built on.
Promote stable logic into a contract
Once you know the dimensions and measures you need, author a contract. This is the moment the logic crosses from draft to governed.
Don't treat exploration as a delivery mechanism. Applications and dashboards consume contracts, not exploration queries. Exploration is where logic gets discovered; consumption is where approved logic gets used.