Find answers from the community

t
tenz
Offline, last seen 3 months ago
Joined September 25, 2024
Hey yall, i'm trying to understand the behavior of deleting a document when the vector store index is an Opensearch one. It seems like based on the vector store module guide, deleting a document makes a call to the VectorStoreIndex's implementation of delete(). For Opensearch, delete() calls self._client.delete_doc_id(ref_doc_id). This makes a call to the opensearch client to delete using id=doc_id. .

The issue i'm seeing is that the id of a document in opensearch indexed by llama-index is the node_id , not the document id. Shouldn't the delete_doc_id delete using a filter on doc_id instead?

as a comparison, chromadb's VectorStoreIndex impl does a delete by filtering on document_id properly here.
1 comment
t
I have a question regarding the use of a global lock in this llama-index guide to support concurrent reads/writes. Does using llama-index's wrappers for remote vector DBs actually require locking? or is this used in the example only because of the use of SimpleDirectoryReader as the vector store?
3 comments
L
t