Zebra AI offers multiple ways to bring your cloud‑hosted Power BI semantic models data into stories. Depending on your workspace and dataset, you’ll use either the XMLA endpoint (Premium/PPU) or the standard DAX API (non‑Premium). Zebra AI detects what’s available and routes you automatically.
When you’ll see each option
- XMLA endpoint: Available for Premium/PPU workspaces. Opens a model‑aware selection UI which builds the corresponding DAX query for retrieval automatically.
- DAX API: Used when XMLA isn’t available. Opens a lightweight DAX query experience, where you’ll have to input a DAX query manually.
XMLA (Premium/PPU): model‑aware, larger extracts
- Selection UI: Browse tables and columns from your dataset model, choose aggregations where relevant, and run immediately.
- AI suggestions: Click “Suggest columns” to get a compact, high‑signal set (prioritizes one core datetime, favors descriptive fields, includes a few key measures).
- Scale and experience:
- Streams results with progress and cancellation.
- Designed for bigger pulls than DAX.
- Limit: up to 10,000,000 values per query.
- Best for: curated extracts from governed models, faster iteration on larger selections.
DAX (non‑Premium): quick queries with API limits
Key Power BI DAX API constraints to keep in mind:
- Not supported: datasets hosted in Azure Analysis Services, or live connection to on‑prem Azure Analysis Services models.
- One query per API call; one table request per query.
- Result size caps:
- Max 100,000 rows OR 1,000,000 values per query (whichever hits first).
- Practical example: 5 columns → up to 100k rows; 20 columns → up to 50k rows (1M ÷ 20).
- Max 15 MB per query; once exceeded, the current row completes and no more rows are written.
- Authentication:
- Service Principals require tenant setting “Allow service principals to use Power BI APIs.”
- Service Principals aren’t supported for datasets with RLS or with SSO enabled.
- Query type: Only DAX queries; MDX, INFO functions, and DMV queries aren’t supported.
Tip: With DAX, keep queries targeted—select only the columns you need, aggregate early, and consider multiple small pulls if necessary.
Security, privacy, and data handling
- Credentials: Access tokens are used only to execute your requests and are not stored in our database.
- Session cleanup: Zebra AI disconnects and discards any temporarily retrieved data when the session ends.
- Safety: Zebra AI performs read‑only operations against your Power BI data.
Typical workflow
- Pick workspace and dataset; Zebra AI opens XMLA (Premium) or DAX (non‑Premium).
- XMLA: optionally “Suggest columns,” refine selection, and run. DAX: write/run a concise query.
- Zebra AI streams results with progress and loads them into your story.
- Save your dataset selection to continue the same story in later sessions without re‑configuring.