We are optimizing the embedding index and updating our LLM inference cluster to improve speed by 40%. All ongoing chat sessions will safely reconnect.