Re: [I] Should we explore DiskANN for aKNN vector search? [lucene]

["weizijun (via GitHub)" <gi...@apache.org>]

weizijun commented on issue #12615:
URL: https://github.com/apache/lucene/issues/12615#issuecomment-2511299239

   hi, all:
       Is there any latest progress on Lucene's diskann? We found that in the RAG scenario, the document data volume is very large and all of it is stored in memory, which consumes a lot of resources. SQ is used to reduce memory overhead, but excessive scalar quantization will lead to a decrease in recall rate. The overhead of knn query accounts for a small proportion of the entire RAG process. Users expect to have a disk-based knn query solution, where the query performance does not decrease significantly and the vector data does not need to be stored in memory.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org