Hacker News new | ask | show | jobs
by klft 1032 days ago
Could you provide more details about your approach? Is the context store based in vector-embeddings?
1 comments

I'd be delighted to assist you with your inquiries. Indeed, the context store interacts with vector databases to retrieve samples based on vector embeddings and cosine similarity.
In your experience, in general when comparing vector embeddings, what layer of the net works best? Are the embeddings from the first hidden layer, the second, the last, etc?