I've been looking into how I can add AI metadata to a semantic model programatically. This allows the push back of AI instruction metadata back to the users in a data ingestable process, notebook and pipeline, bypassing the Fabric 'Prep data for AI' user interface, which you may not want to give to users for iterating through instruction sets.
This works well for some metadata (like table and column descriptions), but does not allow for AI instructions to be written to the model progamatically. I can edit the culture tmdl file via GIT, but when updating back into Fabric from GIT, the AI instructions come back blank.
It would be great to have a REST API endpoint for writing Prep Data for AI configuration.
Cheers,
Kev
No CommentsBe the first to comment
