take 9seconds to search references during chat

[bjwswang]

@ggservice007 Please provide more details about this problem

[ggservice007]

how to trace

kubectl logs -n kubeagi-system arcadia-fastchat-65dbfd7dfc-457c5 -c fastchat-api-server

step

step 1

embedding

step 2

query the pg vector

This step takes about 7 ~ 9 seconds.

step 3

invoke llm to generate the answer