Vectorized Search support similarity score?

[iangithub]

HI all,

just two questions .

(1)
Does Vectorized Search support setting the mini threshold of similarity score?
Currently I only see that I can configure the number of top
var searchResult = await collection.VectorizedSearchAsync(searchVector, new() { Top = 3 });

(2)
At this stage, for integrating Azure AI Search, is it recommended to use the original Semantic Kernel Memory or the new Semantic Kernel Vector Store?

[westey-m]

Hi @iangithub, thanks for reaching out.

(1)
We do not support this at the moment, but we are currently working on a design decision document for it. It has challenges when building into the abstraction since not all DBs support this functionality and the threshold value ranges vary significantly by the type of distance function used.

Follow this work item for updates and a link to the design doc pr when it's ready: #9566

(2)
We recommend the new Vector Store abstractions. We will not be extending the memory store abstractions further, since the design has limitations which prevent us from adding all the functionality needed.

For more information see https://learn.microsoft.com/semantic-kernel/concepts/vector-store-connectors/memory-stores