Use vector search for semantic search in Enterprise Search / App Search

[sebastianboelling (Sebastian Bölling)]

Hi,

I am trying to implement a vector search with App Search. I read this useful documentation how to ingest vectors and ML/NLP generated content with ML inference pipeline.

Document enrichment with ML | Elastic Enterprise Search documentation [8.6] | Elastic

But how to use that additonal vectorized fields (e.g. for title, description, body) within Search API oder Search UI (with App Search or Elasticsearch connector)? The document describes primary the ingest.

Best regards

Sebastian

[joemcelroy (Joseph McElroy)]

You could use elasticsearch with the elasticsearch connector + search ui to perform a KNN search. You can do this by overriding the default query that is used Elasticsearch Connector | Elastic docs and using the KNN functionality in _search Search API | Elasticsearch Guide [8.6] | Elastic

Its a little bit of a workaround solution atm and there are upcoming work that makes this experience so much better with Elasticsearch and Search UI. Using Elasticsearch directly, you should be in a good place to take advantage of that when its released

Joe

[sebastianboelling (Sebastian Bölling)]

Hi @joemcelroy,

Yes, that sounds like a viable workaround. I will test that.

Of course, it would be great if there was a more direct integration with Enterprise Search in one of the next releases. Semantic Search is one of the most important functions for our customers.

Thanks very much.

Sebastian