how does one utilize

distance_threshold

with the automatic embeddings in a hybrid search?

So if you find that you're getting too many irrelevant results with semantic search, you can use the

distance_threshold

parameter in

vector_query

to pick an appropriate threshold that works for your use case. Some docs

thanks but I was asking specifically how to add distance_threshold with the automatic embedding generation

The example in the documentation is for the automatic embedding models

not sure how we’re getting our wires crossed here - I want to leverage the automatic embedding generation, including the search term. I don’t want to be in the business of sending TS an embedding for the query. the example you linked to requires me to do so

let search_parameters = {
  'q'                          : 'chair',
  'query_by'                   : 'embedding,product_name',
  'vector_query'               : 'embedding:([], distance_threshold:0.30)'
}

client.collections('products').documents().search(search_parameters)

The example showcased in the docs passes no embeddings to the search query. It also queries by the embedding field, which is only available on automatic embedding generation

I see. in this case I think the documentation could be a lot clearer. something along the lines of letting the user know that the empty array is valid and required if one wants to set distance_threshold with automatic embeddings

There's mention of querying by auto embedded fields: https://typesense.org/docs/29.0/api/vector-search.html#semantic-search But yes, we'll probably add a small section mentioning how to query by an auto embedded field and manually passing in embeddings as well.

well this particular use case is changing the distance while not passing embeddings. that deserves distinction imho