semantic link cross platform

[smpa01]

Is sempy/semantic-link library useable through databricks. I can connect to fabric one lake in data bricks through SAS.

Is there any way to install semantic-link in databricks and use from there?

[lbendlin]

should be possible. At the end of the day sempy is just a wrapper around the Fabric REST API calls.

Notebook-scoped Python libraries | Databricks Documentation

[smpa01]

"At the end of the day sempy is just a wrapper around the Fabric REST API calls" - At a first glance, it appears so. But I think sempy is more than just that in a sense that it does not have the same limitation of 

/executeQueries . So it has an edge over REST API.

AFAIK from documentation, it appears

read_write_pbi_sempy > read_write_pbi_spark

read-write-power-bi-spark (same as REST API) 

read-write-power-bi-sempy 

Previous Link on Sempy vs Spark for PBI 

I am not interested in config the REST API in databricks, rather to find a way (if possible) to install and execute sempy directly from databricks notebook.

[v-veshwara-msft]

Hi @smpa01 ,
Thnaks for using Microsoft Fabric Community, 

Currently, SemPy is tailored specifically for Microsoft Fabric and does not natively support Databricks. This means that direct installation and utilization of SemPy within Databricks environments are not supported.

From Microsoft Documentation:

​

Reference: Semantic Link Overview for Python | Microsoft Learn

However, alternative approaches exist to bridge the gap between Databricks and Power BI datasets: Semantic-link on Databricks - Databricks Community - 74511

Hope this helps. Please reach out for further assistance.
Please consider marking the helpful reply as "Accepted Solution" to help the other members find it more quickly.

Thank you.

[smpa01 · Accepted solution]

Thanks for the confirmation. I really don't need to use semantic link but I do need query the semantic layer in databricks from fabric.

The true alternative of semantic link is executeQueries Endpoint of the rest api. But the the response returned could be partial dataset as well as AFAIK, and a fetch inside do-while loop needs to be engineered, which is easy to do if the dataset has date parameters or integers, else need to be creatively engineered. I can live with that fir the moment. But I hope there is a better databricks alternative for semantic layer.

[v-veshwara-msft]

Thanks for the follow-up.

You're right about the executeQueries endpoint. Handling partial responses and building out a fetch loop does introduce extra effort, especially when the dataset lacks simple parameters like dates or integers. It’s good to know you’ve found a workable approach for now.

If there are any changes or new capabilities for accessing the semantic layer directly from Databricks, will share an update here.

In the meantime, if any of the responses above were helpful, would you mind marking them as a solution? It helps others who might be looking for similar information.

Best,
Vinay.