Can we also expect
Can we also expect
1. Metadata filters without creating indexes for each
2. Create dynamic vector indexes programatically without the manual bindings
5 Replies
Unknown User•2w ago
Message Not Public
Sign In & Join Server To View
i can live without the #1 but i really need the #2 as i am making a AI agent as a service platform where each user will store their data that i will convert to embeddings and store, but due to current limit i cannot offer unlimited vectors, i know i haven't hit the limit yet but i want to make sure that i can create index on the fly and access them using the env.INDEX_NAME instance
and related to #1, it would be nice with something like upstash, directly we can filter the metadata without extra index creation step, its very straight forward
Unknown User•2w ago
Message Not Public
Sign In & Join Server To View
Just as a stop-gap, you can always use the API to create/query the indexes. I’ve not done tons of testing, but performance-wise it seems pretty similar vs a binding (which I was surprised about - and needed it for a similar use case to yours)
I want to be able to pass multiple index names in an sdk to query data and get data from all
Kind of this
env.MY_DEFAULT_INDEX.query(query_embeddings, {
additional_indexes : […]
})
This can scale to infinite embeddings my users will be able to store
I dont have to query each and every index separately where 1 user data is stored
As to give really unlimited i will be sharing data for all my users across multiple index and not 1 index per user