Slow text search

Support
1

Description

hi Weaviate Community/Team!
I am testing Weaviate text search (bm25)
I use the following code to search

collection = client.collections.get(collection_name)
collection.query.bm25(
        query=text,
        limit=50
    ).objects

for collection with 1 million documents (~ 2000 chars each), the text search works very slowly (8 sec), is there any way to make it faster?

Server Setup Information

  • Weaviate Server Version: 1.25.0
  • Deployment Method: k8s
  • 2 Nodes
  • Client Language and Version: Python client
  • Multitenancy: No
2

Hi @Alibek_Sailanbayev !

Welcome to our community! :hugs:

How is the memory and cpu usage of those pods?

Does the semantic search also experience this slowliness too?

Thanks!

3

Semantic (vector) search is fast ~0.05 sec

4

Ok, so for the text search, you are passing 2000 characters? as a query?

That’s a lot of keywords do search for :thinking:

if using smaller queries, does it get slow too?

Thanks!