If im using propertygraphindex with neo4j and weaviage and i want to store the node embeddings in both but only caclaute then in weaviate (using t2vectransf), how do i setup the embedding pipeline. Should i add to the weaviate index then retrieve embeddings from weaviage then add them to the nodes for the propertygraphindex with ember nodes as false?
Thanks. On the question, id probably invert it, why wouldnt we do that? We’ve been using weaviate in production with collections using a default embedding module for a while now and it works well (faster and more reliable than a 3rd party api) but want to move to a knowledge graph based approach and loving your new index. Slowly learning that llama index doesn’t love weaviate’s modules
Yea in general, llamaindex calculates embeddings and inserts those embeddings into the vector store, rather than letting the vector store embed it 👀 (that way you can use nearly any embedding model, openai, huggingface, cohere, etc.)
