Woukd you be able to make a similar video but instead of document loader, we are using hugging face dataset. The same querying and upserting pinecone and everything into a openai chatbot. Ive been scouring everywhere and you put out the most value. Keep it up man!
@extremehealthradio Жыл бұрын
I've been searching forever to try to find something like your video! I've got hundreds of PDFs, mp3s and URLs that I would like to search to get answers from. With mp3s I'd have to get them transcribed. Can I upload a zip file to bubble or put them all on dropbox and store there for Bubble to connect with?
@BlankSlateLabs Жыл бұрын
hey! just responded via your email. You'd either have to upload the files one at a time or upload them somewhere else and add a link(s) to the record in Bubble. It does look like Bubble has a plugin for using Box to store files (manual.bubble.io/core-resources/bubble-made-plugins/box), though I personally have never used it, so can't comment too much on its use.
@jvanh8926 Жыл бұрын
Thanks for your video! Is pinecone the only option though?
@BlankSlateLabs5 ай бұрын
Not necessarily. There are three types of options. 1. Fully Hosted Vector Databases - Pinecone - OpenAI's vector stores (as part of the Assistants v2 API) simpler to use, but less control over how you store and retrieve data. 2. Vector data types in hosted backend services (using PGVector - an add on to Postgres) - Supabase (with PGVector enabled) - Xano (using their vector data type) 3. Self-hosted vector stores - Chroma
@shoshi475 Жыл бұрын
Hi there, I am impressed by your work and channel. How would I be able to contact and hop on a call with you?
@BlankSlateLabs Жыл бұрын
Sure! Feel free to reach out to me via email: hello [at] blankslate-labs.com