Hi I am looking for some advice regarding the following.
I have two Class Products and Queries. Some times I need to search just between products. Other times I need to search between products and queries.
For each product I would like to calculate the Top N nearest search queries. In doing so I am wondering what is the better option of these:
(a) Batch pull all the product vectors into memory then for each vector preform a near_vector search on the queries class.
(b) Create a ProductQueries Class with boolean filter is_product . Then preform a near_object search for each product ID filtering results with is_product=False.
I see potenial speed up with option (b) because it doesn’t require having to retrieve the vector and submit it in the api request. However, I wonder if the trade off having queries and products both in the class and having to filter out products each time has any more cost.
Any advice that could be offered would be greatly appreciated
I like option b more, as having batching pulling products and doing N+ queries will probably be worse than a single query, that will be filtered out internally.