Hey Guys, I have a client that is looking at using OpenAI through a custom app (Glide) to analyse large documents…How could one approach this?
If you’re referring to that as a large document, I’m not sure if these approaches work:
Somehow taking all the text from that doc and feed that in the prompt to OpenAI. It would exceed the tokens limit.
Upload the file to Glide and give the link to OpenAI to “browse”. Maybe something worth trying, but they have recently stopped beta testing for the Bing browsing feature. I wonder if they stop the ability for gpt-3.5 and gpt-4 to stop being able to read external links altogether.
The best approach on the market is using vector embeddings, but that is not something that is built directly into Glide.
Any updates on this topic on your side?
I’m trying to implement something similar to create a database from archives in PDFs. I have a set of key properties I need to extract from each PDF to build my database. Then I’ll be able to create ML model hopefully.
Thanks for any tips or idea on how to approach. Might not even need Glide for the moment but rather an automation workflow via n8n or make