DXXS
DXXS
Hmm; well I halved the limited parameters available 'chunksize' and 'overlap', and got somewhat closer results on that one document (though it somehow split it right between 'DEFINITIONS' and 'means',...
Ah! it looks to me as though, if this project were to add something like this, it might help! "A Framework for Analyzing and Improving Content-Based Chunking Algorithms" : https://www.hpl.hp.com/techreports/2005/HPL-2005-30R1.pdf...
FWIW, I was just granted GPT4 developer status (though apparently the API is only accessing the chat model, and something like a davinci-004 isn't available yet). 'Chatting' with my documents...
I don't consider the issue to be resolved; minor improvements can be achieved by the given modifications, but I haven't seen fixes for the the 'broad spans of data' portion...
I've seen no responses from the developers that these issues have been substantially fixed; why don't you point us to the evidence? Managing 'backlog' by claiming that issues have been...
I found that deleting the extant pinecone vectors before re-ingesting solved most issues where it seemed to think that the original docs were still available. However, it seems to be...
Then too, I had originally hoped I could just use symbolic links, but apparently it somehow doesn't recognize those as valid, nor OSX alias's either!?
Bot should probably consult developers to verify if 'still relevant', as I haven't seen mention of associated fixes yet.
Got this here too, with gs 10.0