Open faisalron opened 2 weeks ago
Good point! Would you like to work on this?
Hi. I think there is not even need to make that one available as an option but just make it not indexable. At the end of the day the entire thing just works through id retrieving
Hi Everyone,
I am using VectorSearchVectorStoreDatastore as my vector store for building RAG system. Turns out that I can't put metadata that exceeds 1500 bytes to the vector store. I think it's because the limitation of the Datastore used as the backend.
I think we can relax this limitation by passing exclude_from_indexes argument to the following line of code: https://github.com/langchain-ai/langchain-google/blob/main/libs/vertexai/langchain_google_vertexai/vectorstores/document_storage.py#L191
Reference Documentation: https://cloud.google.com/python/docs/reference/datastore/latest/client#entitykeynone-excludefromindexes
I can ingest the same data with VectorSearchVectorStore. But with latency constraint, it would be good to add this feature.