Hello. We're considering migrating from auto-embedding to externally generated embedding for vector search and, based on the docs, we'll lose the ability to perform hybrid searches. Is this correct? I'm assuming we can do a multi-search with both keyword and vector searches, then merge and rerank. Do you have any guidelines/recommendation on how to do achieve this? We're using typesense cloud v29.

You can pass in a normal

vector_query

parameter with your externally generated embeddings, as you would when using semantic search by itself

If you are okay with the doing the multi_search. The simpler which I could think of Just take ranks of document in both vector search and keyword search and compute the

K = rank of document in keyword search
S = rank of document in semantic search

rank_fusion_score = 0.7 * K + 0.3 * S

and sort this in ascending order. You will have results close to hybrid search results. It is more accurate the more number of results you take from both searches and rerank.

I completely missed this example in the docs. We can combine a keyword search and vector search in the same query. Thanks @Fanis Tharropoulos!