Any way to create a "shortcut" to a semantic model?

[dbeavon3]

A fellow PBI developer is building pyspark solutions.  He is very happy with the "shortcut" capabilities which allow you to connect to various types of data sources, including remote blob storage.  Here is more information about that from Microsoft:

https://learn.microsoft.com/en-us/fabric/onelake/onelake-shortcuts

Has anyone tested the ability to create a shortcut to a remote semantic model (ie. dataset)?  Are there any special security considerations that are specific to the creation of lakehouse shortcuts?  If a user has been given rights to access a semantic model via sempy, then would they also be able to access it with a "shortcut"?  Are there any problems that arise when creating a shortcut to a dataset that lives in another team's workspace?  Are there problems when connecting to a remote dataset in a different Azure region?

Any information would be appreciated.  If you look at the link, the list of sources for shortcuts does NOT seem to include semantic models.  Is there a way to connect to semantic models (possibly via a "preview" feature)?  What would be the ultimate benefit of creating a shortcut to a semantic model?  (ie. as opposed to alternate types of connectivity like using "sempy" or "PQ").

[v-menakakota]

Hi @dbeavon3 ,

May I ask if you have resolved this issue?. Please let me know if you have any further questions or if you'd like to discuss this further.

Thank you and Regards,
Menaka.

[StrategicSavvy]

I think what @lbendlin  mentioned is right. If you want to create shortcut you can materliaze data in the same WS where is your semantic model by creating pipeline and call power bi api which sink it in lakehouse and then create shourtcut. A little. bit overkill but if this something that you need that might work.

[lbendlin]

A semantic model is a SSAS Tabular source. Not as complex as an OLAP cube, but still more complex than a regular database or file/blob storage. 

I'd say a live connection to a semantic model is akin to a shortcut?

[dbeavon3]

Hi @lbendlin 

Just a follow-up.

I found out what the users were referring to.  It wasn't straightforward.  There is a preview feature of a semantic models that allows you to enable "onelake integration".

The purpose of this appears to be so that it will poop out a bunch of deltatable files (parquet), as a byproduct of the refresh operation.

As long as users are admins/members/contributors (ie administrative users) then it works.  I'm obviously not happy about making everyone administrative users, just so they can get data that they can otherwise get using DAX and MDX.  I guess this may be one of the reasons it is still in preview.

[lbendlin]

You mean this one? Learn about Microsoft OneLake Delta table integration in Power BI and Microsoft Fabric - Power BI | ...

Tail wagging the dog much?

[dbeavon3]

Yes, that is the new feature they were looking to use.

Instead of this, I proposed that the users should rely on the sempy api in python.  That is a much better way to get data from a dataset and move it to blob storage.  I can definitely support sempy a lot better than allowing everyone to go after the pooped-out parquet. 

The pooped-out parquet is especially unfriendly when the target audience is data scientists and analysts.  These folks shouldn't need to mess with the surrogate keys and hidden tables that are holding the dataset together on the inside.

The worst part about the feature is the permissions requirement.  As it works now, this feature encourages everyone to request administrative access in all the workspaces where their data comes from:

Learn about Microsoft OneLake Delta table integration in Power BI and Microsoft Fabric - Power BI | ...

I'm guessing this is a valid option for small PBI organizations of five people - where everyone is already an administrator to start with.  But in an organization of dozens or hundreds of PBI users, it doesn't make sense to hand out the "contributor" role to all the users of a data source.

[v-menakakota · Accepted solution]

Hi  @dbeavon3  ,

Thank you for sharing your insights! We appreciate your detailed feedback on the OneLake Delta table integration and the comparison with sempy for data extraction.

Your approach of leveraging sempy for moving data from a dataset to blob storage is indeed a well-supported and flexible solution, especially for Python-based workflows. It allows for direct queries to the semantic model using DAX and MDX, without requiring administrative permissions in multiple workspaces.

Regarding the OneLake Delta table integration, we understand your concerns about the permissions model. Currently, access to OneLake-integrated semantic models requires Admin, Member, or Contributor roles, which may not be feasible in larger organizations with strict access controls. We recognize that this could lead to challenges in governance and security.

For more details on OneLake Delta table integration and its security considerations, please refer to:
https://learn.microsoft.com/en-us/fabric/onelake/security/data-access-control-model

We highly value your feedback on this feature and encourage you to share your thoughts through the Fabric Ideas forum to help shape future improvements:

https://ideas.fabric.microsoft.com/

Please let us know if you need any further clarification or support. We appreciate your engagement in the Microsoft Fabric community!

[v-menakakota]

Hi @dbeavon3 ,

Could you please confirm if you've submitted this as an idea in the Ideas Forum? If so, sharing the link here would be helpful for other community members who may have similar feedback.

If we don’t hear back, we’ll go ahead and close this thread. For any further discussions or questions, please start a new thread in the Microsoft Fabric Community Forum  we’ll be happy to assist.

Thank you for being part of the Microsoft Fabric Community.