There’s nothing technically stopping us from adding image classification it’s just going to be really expensive. The current classifier is mostly a proof of concept for us to classify posts to train a cheaper classifier model that will eventually take over.
In order to support the scale of filter.nostr.wine aggregator we need to reduce per note costs significantly.
what’s the architecture look like? are you using elastic’s builtin NLP stuff or something diy?
Please Login to reply.
No replies yet.
