Quick question I’ve been using Typesense Cloud on...
# community-help
s
Quick question I’ve been using Typesense Cloud on a 2GB RAM instance for the past month, with average usage around 25%. I’m now trying to integrate image search by enabling auto-embeddings using the default CLIP model. However, the embedding generation keeps failing due to memory constraints. Could you please suggest the most cost-effective setup that would allow me to run image search reliably, without a major increase in cost?
e
I'm surprised that runs on 2GB. I did embeddings with a 4GB self-hosted and it really struggled, but it did finish (3000-ish records on that collection)
a
Hey @Sharnam Singhwal, You'd want to make sure your RAM holds your entire dataset in it. You can check the system requirements for vector search here: https://typesense.org/docs/guide/system-requirements.html#for-vector-search