search

infiniflow / ragflow

RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding.
https://ragflow.io
Apache License 2.0
23.62k stars 2.32k forks source link

[Question]: HTTP API Retrieve chunks efficiency #3260

Open lvyoudashuju opened 2 weeks ago

lvyoudashuju commented 2 weeks ago

Describe your problem

There is only one node deployed now. If multiple nodes are added, will the retrieval efficiency be improved? Or is there any way you can improve the efficiency of retrieval? 企业微信截图_17309624306746

KevinHuSh commented 2 weeks ago

Re-rank model is very time consuming. I suggest not to use re-rank model since it's low in terms of ROI.

lvyoudashuju commented 2 weeks ago

Reordering is required for accurate retrieval.

Match-Yang commented 6 days ago

Re-rank model is very time consuming. I suggest not to use re-rank model since it's low in terms of ROI.

Even if I don't use the re-rank model, it's still slow. Are there any optimization methods? Is it related to the block size of the embedding or the embedding model?

KevinHuSh commented 4 days ago

Click on it and you can see the elapsed time of every step of retrieval. Paste that screen shot here if it's convenient for you.