Is there a standard way to query using multiple vectors? I’m having pretty good success simply averaging my query vectors (numpy.mean()) before querying (against a cosine index), but it would be nice if this is built in to the query API. I’m no expert but I decided to try simply averaging the vectors from reading this issue: Search neighbors of mulitiple query vectors based on the average distance? · Issue #1257 · facebookresearch/faiss · GitHub
Hello Beatty,
Welcome to Pinecone! Great question. It is typical to take the average of the vectors before submitting them as a query to Pinecone just as you’ve done. We are taking your input and passing it to the product team to consider for our roadmap. Thanks for the input!
Pinecone
Hello, Any Update on this? Query using Multiple vectors much needed for efficient search.
@shardul We chose to deprecate batch queries because, in many cases, an asynchronous approach actually proved to be more efficient.
Have you tested performing the queries asynchronously? If so, what kind of performance were you observing? Additionally, you can try averaging the vectors and running one query, as mentioned above.