hey all. the documentation for hybrid search is a bit confusing. https://typesense.org/docs/29.0/api/vector-search.html#semantic-search

When using auto-embedding, you can set

query_by

to a list of both regular fields and auto-embedding fields, to do a hybrid search on multiple fields.

If you are populating the embedding field externally, without using auto-embedding, you can still do a hybrid search by passing the embedding of the query string manually via the

vector_query

parameter and omitting the embedding field from

query_by

.

does this mean that when we don’t use auto embedding, we have to create embeddings for the search term we want to search by, and include that embedding in the query payload? does this also mean that we can’t use the embedding field in the query_by? it would seem odd that we can’t provide the embeddings ourselves (so that we have control, retry, etc on our end) and also tell typesense what model we want to use

When you don't use auto embed, we won't know what model was used to embed the documents, so the query embedding has to also be passed in via

vector_query

parameter.

gotcha. it would be great if typesense could accept the model name that explained what the format that the embeddings provided were in, so that the vectory_query didn’t have to be provided. but probably not high demand for that

Yeah, in general if people are having infrastructure to do the embedding of docs, they just do the query as well, but happy to accept this as a feature request! Please create a github issue.