I’ve been using Azure Search for months now, using an Indexer to load data in from an Azure SQL database. I think use the Azure Open AI functions with the cognitive search extension, to provide essentially a “Bring-You-Own-Data” chatbot experience.

This has been working really well, but I was asked what embedding the process uses, so I tried to find out. I don’t know where to look, and am even having doubts now about whether I am using one at all – I assumed loading the data in via the indexer would apply a default embedding to create the vectors and then it all just works, without much effort. So which is it? Is there a default embedding, or am I not using an embedding at all and it’s just using full text search or something – although the results have been excellent so I can’t imagine it getting much better…