Wow. I can't believe it's live. I waited forever for Azure Code Signing (earlier name of this) to emerge in preview after the deprecation of the Device Guard Signing Service left us needing this, but then it never came for years and I eventually lost sight of it. Glad it's here at last I guess.

FWIW, in the December shipmas post about o3 they said o3-mini was scheduled for release in January, but I take your broader point

Understood. You could start here: github.com/microsoft/ke.... We normally build with Azure AI Search which includes services like Document Intelligence

Finding the right chunk size (and the related delimiter(s) to parse on) can take some experimentation. Ultimately there is a whole art to this that is common to the world of traditional indexing technologies, but your data may submit well to something cheap/cheerful like this

There are frameworks/services to help with it if you have the appetite to dive in. Otherwise you can find delimiters to parse on (I find HTML is more reliable than MD), treat each as your chunk, verify those chunks will fit with your embeddings service, then store those embeddings in your vector DB

This would get messy with overlap, but in the first instance for something like this I would test without overlap, since the section should have some completeness/coherence to it