Dimensions
Dimensions are the ways a metric can be split, such as by channel or by month.
A metric is the number, such as return on ad spend (ROAS) or net sales; a dimension is one way to split it. A semantic layer defines both, and a dashboard or an agent asks for a metric by name, with the dimensions, filters, and dates it needs.
Under the instructions dbt ships with its connector, the agent first asks for the metrics and dimensions it can use, then queries a metric by name, such as total_sales grouped by month. A split nobody defined is not available to it, so the agent says what is available and suggests a close question it can answer.
Example
In the example brand, requests for platform ROAS by ad set and audience keep arriving in the question log. The data team adds ad set and audience as ways to split platform ROAS. It answers a rare request by hand.