DirectLake query mode based on lakehouse views

Hi All -We aim to develop a semantic model using lakehouse views instead of tables. Are there any potential fallbacks or considerations we should keep in mind? Since we're not incorporating all columns from the lakehouse tables into the semantic model, our approach is to build it based on views.

One option is to reprocess the data, limiting the columns. However, this includes additional processing and setup. Do you have any suggestions or alternatives to streamline this process?

Thanks,

Abhiram