I have created thunder, a fast kv database for ai workload. Our main motive was to handle large datasizes (2mb-512mb+) with lighning speed and memory should not be our constraint. We use memory for buffer only and we use the raw speed of nvme to directly stream the data. It is best used for memory, large chats, other stuff. I am happy to answer all your questions. https://thunder-db.com/
Top comments (0)