Warehouse as source for Ontology

[Asmita_27]

Hi Team,
I’m working with Ontology in Microsoft Fabric and want to use a Data Warehouse as the underlying source instead of a Lakehouse.
Apart from using shortcuts, is there a recommended or supported approach to directly use Warehouse data in Ontology?
I also tried creating the Ontology via a Semantic Model built on top of the Warehouse, but the ontology bindings are not getting created/mapped correctly (relationships and entities are not resolving as expected).
Is this a known limitation with Warehouse‑based semantic models, or am I missing any configuration steps or best practices?
Any guidance or workaround would be appreciated.
Thanks.

[arabalca]

Hi @Asmita_27 ,

For now, this is where things stand…

I would recommend opening a ticket with Microsoft, as it’s important for them to be aware of this issue.

You could also try using a Lakehouse-based semantic model, although you might run into the same problem.

For now, we’ll have to wait.

Thank you very much for your response. If you think this answers your question, please mark the solution as accepted so other users can benefit from it.

[arabalca · Accepted solution]

Hi @Asmita_27 ,

When working with bindings in Microsoft Fabric, it’s important to understand which sources are currently supported.

According to the official documentation ( https://learn.microsoft.com/es-es/fabric/iq/ontology/overview#data-binding ), the valid sources are:

• Lakehouse (Delta tables in OneLake)
• Eventhouse / KQL Database
• Power BI Semantic Model (used to generate the initial structure)

The Data Warehouse is not a supported direct binding source.

In addition, using a semantic model based on a Warehouse is not fully reliable at the moment. Relationships are imported incompletely, and there are column mapping issues that can break the bindings.

Let’s keep in mind that this feature is still in preview, so this kind of behavior can be expected.

That said, if you run into this scenario, I would encourage you to share the error, as it may help other users.

In the meantime, there is a workaround that works.

By creating a shortcut in a Lakehouse pointing to the Warehouse tables, you can bind from the Lakehouse without issues. I’ve tested it and it works correctly.

It’s not a critical limitation, but it’s something to keep in mind when designing your solution.

[Asmita_27]

Hi @arabalca,

Yes, thanks for the explanation. We were already aware of the supported sources, which is why we tried the Lakehouse shortcuts earlier.
Our main challenge is that we already have multiple semantic models built on top of the Warehouse, and a large volume of data resides there. Creating shortcuts for all tables isn’t ideal, so we were exploring whether there’s any other supported option or best practice for this scenario.