In theory it should be possible to connect the actions of a custom ChatGPT GPT to a Pinecone API in order to access its vector database capabilities. However I haven't seen anyone talking about this. Is there a reason or is it simply that no one has tried?
#Connecting a ChatGPT GPT action to vector database API?
night lily
worn turtle
In theory it should be possible to connect the actions of a custom ChatGPT GPT t...
I think nobody's tried. I've had success connecting GPT directly to SQL servers in the past. What're you trying to build anyway?
From my experience I've found that it's generally better to give GPT more targeted tools (when applicable) than to allow it general access to a db. Seems to cut down on confusion
night lily
Connecting a ChatGPT GPT action to vector database API?
night lily
I think nobody's tried. I've had success connecting GPT directly to SQL servers ...
I have about 30 compliance documents from a merger and my team is struggling to understand all of the nuances of the legal issues concerning the merger and we getting bogged down in delays to development. I want to upload those docs to Pinecone to turn into a searchable vector database for a CharGPT GPT to consult.
But I'm struggling to understand how I would implement this within ChatGPT as a custom action within a GPT. I have it working locally in MemGPT but the thing is I am not the person who needs this. I have to either get department managers to read all 30 documents (lol never going to happen) or give them a GPT that can answer their ceaseless questions.
They can use a custom GPT with a link I give them. But I cannot get them to run their local memGPT nor can I get them to use a memGPT running in a DigitalOcean environment due to the high expense of such a server setup.
worn turtle
I have about 30 compliance documents from a merger and my team is struggling to ...
How many pages are we talking? This sounds like something FAISS could do without much issue. I've got an implementation I did for a hobby project that handles 2 versions of the bible and a couple supplementary books really well. The whole thing runs on a cloud function with like 1GB of memory if I remember right
I'd be happy to share if it'd help
At least from my understanding of Pinecone (which tbh I haven't used a whole lot myself) it's generally best saved for larger projects than a couple books/couple dozen documents worth of embedding
night lily
Like a couple hundred pages per PDF. Dense compliance text.