Bring Power BI models into One
Upload a Power BI model and ask questions in plain English.
If your team's metrics already live in a Power BI semantic model, you don't have to rebuild them to use Databasin One. Bring the model in, and the agent translates its DAX measures into SQL over your lakehouse — so you can ask the same questions in plain English, against the same definitions.
Bring your model in
In the Power BI panel of the chat sidebar, upload a TMSL export of your model — a .tmsl.json file from Power BI, Tabular Editor, or SSMS. Databasin reads the model, maps its tables to your lakehouse tables, and reports what it found (the table and measure counts). Check the box next to the model and Apply to add it to the chat context.
Upload the model's TMSL (.tmsl.json), not a .pbix. Exporting through Tabular Editor captures the measure expressions, which is what makes the translation accurate.
What the agent does with it
Once the model is in context, its measures, calculated columns, and relationships travel with every question. When you ask something, the Power BI skill translates the relevant DAX into SQL:
- It reads each measure's DAX body as the source of truth — not the measure's name, which can be misleading.
- It expands measure references and inlines calculated columns down to physical columns and arithmetic.
- It handles time-intelligence patterns and active vs. inactive relationships (including
USERELATIONSHIP), then runs the SQL against the mapped lakehouse tables. - It cites which measures it used, so you can check the number against your Power BI report.
The result: "what was net revenue by region last quarter?" returns the same figure your Power BI report would — now in conversation, and joinable with everything else in your data context.
When something's missing
The agent is built to be honest rather than confidently wrong:
- Measure with no DAX body. Some tenants return a measure's name and description but withhold its expression. The agent approximates from the name and description over the mapped columns, and tells you it approximated so you can verify.
- Ambiguous table mapping. If a Power BI table could map to more than one lakehouse table, the agent asks you to pick instead of guessing.