Embedded Power BI Notebook Renderer Exploration

[sbatten]

This item covers exploration of what the intersection of Power BI and Notebooks could be.

[sbatten]

Possible Scenarios

1. Embed Pre-existing Full Report or Chart as Output to References
2. Create Report from Kusto Queries using custom language
3. Provider Scenario Load PBIX file as a notebook with data sources and charts as different code cells

References

Power BI Embedded

Power BI REST API

Power BI JS

[sbatten]

Embed Pre-existing Full Report or Chart as Output to References

Value

Can be used as a reference to support an investigation in a Kusto-style notebook or additionally, after investigation, a user might create a report specifically to embed in a published notebook.

Cons

Requires external tooling to actually create the rendered component.

Feasibility

This can be done with existing Power BI APIs. After setting up the login process and possibly a simple sytax for requesting a Power BI report/dashboard (with Intellisense), it would be simple enough to implement.

Create Report from Kusto Queries using custom language

Value

User could create reports on-the-fly while using kusto or other data sources for investigation in the notebook. Using a custom syntax, we could template reports or charts and plugin queries.

Cons

Foreign method of working with cloud-based Power BI reports. Less interop with Power BI desktop.

Feasibility

AFAICT, it just isn't possible to hookup queries to reports when making charts in Power BI using the existing API. One workaround would be to ingest the data locally using a query language, then push a copy of the data manually to a new report.

Provider Scenario Load PBIX file as a notebook with data sources and charts as different code cells

Value

This essentially provides a new mode for working with PBIX files. All changes to the PBIX would be interoperable with Power BI desktop and provides the user a better way of investigating the data while creating a report. Right now working with kusto queries and PBI Desktop seems very disjoint.

Cons

The cons of this approach are primarily in the feasibility mentioned below. Also, since notebooks don't have the same report editing fidelity (e.g. layout of charts), it may feel cumbersome when moving between tools if that option is taken.

Feasibility

This is a fairly expensive endeavor as we would really need to copy a lot of the report editing code from the web client of Power BI into our provider/render. Also, the start of this would be very limited. Each different type of data source would need to be handled and I imagine kusto would be the start.