HomeGuidesSDK ExamplesAnnouncementsCommunity
Guides

Datasets in Max act as a semantic layer that organizes and structures data from connected sources, making it easier for Max’s language models to interpret and respond to user queries. Built on top of connections, datasets describe the underlying data by identifying key metrics, dimensions, and other important attributes. This allows users to interact with the data using natural language, without needing to understand complex database structures or write SQL queries.

Datasets enable flexible data exploration and analysis, allowing Max to guide users through their data by grounding natural language questions in the dataset’s structure. As data is updated, datasets can be refreshed to ensure they contain the most current values, supporting up-to-date insights and responses.

Each Dataset contains top level properties.

Dataset Name: A friendly name for the dataset. This name is typically presented to the language model during chats so be sure to choose a meaningful name.

Data Interval: The frequency of the data in the dataset. Choose from daily, weekly, monthly, quarterly and yearly. This information is often used by the skills at runtime.

Source: This is the table and connection the data is sourced from. It is possible to edit the source if the underlying data has moved to a new location.

Description: This is a description of the dataset. This is presented to the language model during chats so it's important to be clear and informative.

Dataset Date Range: The time range for the dataset. This is presented to the language model while chatting to ensure it is aware of any data limitations.

Query Row Limit: This is the maximum number of rows to return for any query. The default is 25K rows.

Use Database Casing: This enables skills to present dimension values using the casing found in the database. Otherwise the casing will be lowercase.