Hey quick question on loading locally stored knowledge graphs - should I store with storage_context.to_dict() or with persist()? And how do I load the knowledge graph from these files? Recommendations?

Have you seen studies whether (and for which models) it matters if the retrieved data is fed as a user, system or assistant message to OpenAI chat api?

Anybody using the Qdrant vector index? It seems like doc_id is lost on the source nodes when querying

Just saw the SQLStructStoreIndex, does this use embeddings at all?

@jerryjliu0 saw your tweet about the new turbo-instruct model a few days ago, any updates regarding your hunch on its capabilities?

https://status.openai.com/ right now openai is having outages, may explain your errors as well

I've got an issue where I can form a Qdrant index but cannot use the reader, it gives a "name or service not known" error. Seems to come from the Qdrant client. Anyone working with it?

I make a query, and get a response. Response has this list of source nodes. I would like to map these to embeddings. How?

Are the original document names saved in the index?