Forum Discussion
Direct Lake Composite Mode
Hi jaryszek ,
Thanks for getting back.
You are correct - at the moment, when working with Direct Lake + Import composite models, you cannot edit Import tables directly in Power BI Desktop. Adding and managing those tables is only possible through external tools such as Tabular Editor, Fabric Studio, or Semantic Link Labs. Since this feature is still in preview, it is expected that more direct editing support will come once it reaches general availability.
The purpose of having Import tables alongside Direct Lake tables is to give flexibility in your design. Direct Lake is best for very large fact tables because it provides performance without loading data into Vertipaq, but it has limitations such as no calculated columns or Power Query shaping. Import tables, on the other hand, let you include smaller dimension or lookup tables from other sources and extend them with additional modeling features like hierarchies, calculated columns, or custom structures.
From a design point of view, the difference is that you can mix storage modes to balance performance and modeling flexibility. Direct Lake handles scale and speed for large fact data, while Import tables give you the ability to enrich the model with smaller, more structured data.
Hope this helps. Please reach out for further assistance.
Thank you.
Hi jaryszek ,
Thanks for posting in Microsoft Fabric Community and sharing the reference.
As MasonMA suggested, the video walkthrough is useful for seeing how the process works in Tabular Editor.
To clarify, when the documentation says “Paste or create Import tables (via Tabular Editor 2/3 or soon in-product),” it means that after starting a Direct Lake model, you can add additional tables in Import mode into the same semantic model using authoring tools such as Tabular Editor, Fabric Studio, or Semantic Link Labs. Full GUI support in Power BI is still under development, so these tools are the current way to create composite models.
Regarding your question about schema changes, Direct Lake tables will continue to reflect schema updates in OneLake, but Import tables work differently. If the schema of the source changes, you need to update or reprocess those Import tables through one of the supported authoring tools. The copy step is not automatic; once an Import table is created, you can refresh its data, but structural schema changes (like added or removed columns) require re-authoring.
As the Microsoft blog also points out, once a model is created you can open it in the web to make semantic model changes such as measures, relationships, hierarchies, and columns. However, schema refresh of the tables themselves is not yet supported in the web experience.
You may also find this video helpful for another walkthrough of the process: https://www.youtube.com/watch?v=Im7Cmr1il3o
For additional details, you can also check this: Composite Model: Direct Lake + Import - Microsoft Fabric Community
Hope this helps. Please reach out for further assistance.
Thank you.
Thank you.
But I have still question.
Having tables in import mode should give me a possibility to run power qeury on them ?
Or what is the purpose?
What is a difference for power bi design?
Best,
Jacek
Hi jaryszek ,
Thanks for getting back.
You are correct - at the moment, when working with Direct Lake + Import composite models, you cannot edit Import tables directly in Power BI Desktop. Adding and managing those tables is only possible through external tools such as Tabular Editor, Fabric Studio, or Semantic Link Labs. Since this feature is still in preview, it is expected that more direct editing support will come once it reaches general availability.
The purpose of having Import tables alongside Direct Lake tables is to give flexibility in your design. Direct Lake is best for very large fact tables because it provides performance without loading data into Vertipaq, but it has limitations such as no calculated columns or Power Query shaping. Import tables, on the other hand, let you include smaller dimension or lookup tables from other sources and extend them with additional modeling features like hierarchies, calculated columns, or custom structures.
From a design point of view, the difference is that you can mix storage modes to balance performance and modeling flexibility. Direct Lake handles scale and speed for large fact data, while Import tables give you the ability to enrich the model with smaller, more structured data.
Hope this helps. Please reach out for further assistance.
Thank you.