michaelthwan / searchGPT

Grounded search engine (i.e. with source reference) based on LLM / ChatGPT / OpenAI API. It supports web search, file content search etc.
MIT License
621 stars 65 forks source link

Profiler to profile program #32

Closed michaelthwan closed 1 year ago

michaelthwan commented 1 year ago

e.g. I think the 5x scrap website is quite slow

peterhcyuen commented 1 year ago

searchGPT_profile

For reference

karabC commented 1 year ago

If the bottleneck is the embedding service for Semantic search, should we try to setup config to use external provider, e.g. says cost effective GPT text-embedding-ada-002 for trial?

We can still have the sentence-transformers/all-mpnet-base-v2 as the default choice, but externalize the loading to other embedding service endpoint seems making more sense to me.

michaelthwan commented 1 year ago

@karabC possibly~

michaelthwan commented 1 year ago

image

michaelthwan commented 1 year ago

quite fast now. Let me close this.