Content
@
0 reply
20 recasts
20 reactions
Matthew
@matthew
made something new with @kevinoconnell and we’re calling it hubscout.xyz! Hubscout is an embeddings search engine for Farcaster. You can use it to find things semantically on Farcaster. For example, normal search would return nothing for “Greg hot chocolate”, whereas semantic search returns lots of results.
20 replies
30 recasts
131 reactions
welter
@fun
i love this, but curious how difficult it would be to ask something like "what is the most liked cast"
1 reply
0 recast
0 reaction
Gabriel Ayuso
@gabrielayuso.eth
Different approach to answer analytics questions as opposed to content based questions.
1 reply
0 recast
1 reaction
wizard not parzival (shoni)
@alexpaden
I have two databases one focused on semantic speed and the other analytical SQL queries- a fine tuned model builds the query based on user prompt so in the end either is possible but limited by current database query speeds/ability with a 1-3s delay added. Technically I could use the vector plugin for postgres but pinecone is very efficient. As for how to make the model, it's essentially the same process as https://chatgpt.com/g/g-lKnQHXJKS-dune-x-farcaster-gpt which handles far more intense queries than i intend to allow (speedily on dune)
0 reply
0 recast
0 reaction
