AI model

The assistant answers with your own LLM API key — Anthropic (sk-ant-…) or OpenAI (sk-…), detected automatically. Stored locally in config.json.

Power BI connection

Register an app (service principal) in your own Entra tenant, create a scratch Power BI workspace and add the service principal as a Member, then save the details here. Uploaded .pbix files are published into that workspace, read back, and deleted. Stored locally in config.json.

Onboard a Power BI workbook

Two ways in. Both read the model's tables, measures and relationships, extract and profile the data, and generate the reference documents with AI.

From a published report URL (recommended)

Paste a report's URL from powerbi.com (while viewing it). The report is read in place — no upload. The service principal must be a Member of that report's workspace.

Or upload a .pbix file

An Import-mode .pbix with embedded data. It's published to your scratch workspace, read back, then the scratch copy is deleted.

Assistant knowledge

Give the assistant extra context it can use when answering — business definitions, targets, caveats, anything not visible in the data itself.

ⓘ Overall description (below) applies to all dashboards — e.g. "All amounts are in USD; the fiscal year runs Apr–Mar."
Per-dashboard notes — click Notes & files on a dashboard in the list below for context specific to that dashboard.
You can type text, upload a document, or both. Supported file types: .txt, .md, .docx, .pdf (max 5 MB). The text is extracted and handed to the AI alongside the generated reference documents whenever it answers.

Onboarded dashboards

Everything the assistant can currently answer questions about. Use Notes & files to add curated context for a specific dashboard (text or a .txt / .md / .docx / .pdf file).

Loading…