I think the issue is more that when

The community member is using PGVectorStore and is unable to directly get the nodes. They are using vec_store.query(VectorStoreQuery(similarity_top_k=100_000_000)).nodes to fetch all nodes, but the returned nodes have no embeddings despite the database showing the embeddings in the corresponding column.

In the comments, another community member suggests that the implementation is probably not fetching/attaching the embeddings, and that this approach makes responses faster and uses less memory. The original poster asks if there is a way to enable this, or if there are other vector stores that support this, as they are getting a NotImplementedError when using get_nodes() on PGVectorStore.

In the end, the original poster manually fetched the rows and created TextNodes.