AI Cookbook - nextjs-with-flask-server upload fails

I’ve installed the nextjs-with-flask-server from the open ai cookbook (the forum won’t allow me to include a link).

When I attempt to upload a file, either by drag/drop or file upload (Windows) I see the ‘uploading’ message, but the upload is not successful. In Dev Tools Console I see:

Uncaught TypeError: t.drop is not a function
    at content.js:1:38480
    at Object.handleEvent (content.js:1:41982)
(anonymous) @ content.js:1
handleEvent @ content.js:1

Any help will be appreciated.

I discovered there was a mismatch with my pinecone index, so the browse/upload method is working. The drag/drop error remains.
I tried to analyze some php files which I renamed to txt in order to upload them. The responses are very limited, mostly:
“I couldn’t find the answer to that question in your files.”.
It doesn’t have the depth of GPT.

This is my first exposure to the API. Is it a lot different than GPT?

Hey there @mitch3, welcome to the community!

Now, as your first exposure to the API I suggest going over the documentation and getting used to it without pinecone. What are the sizes of the files you are using? if they do not go over the API request size limit, you wouldn’t even need to use pinecone. In my experience, using vector databases can be quite tricky and can often end up and with subpar results if not done correctly. As far as I know, OpenAI does not provide any documentation on using vector databases. At the end of the day, I feel like I move faster by just thinking of creative ways to get what I’m looking for without going over the token limitation, as deep down it feels like the quality of the response is somehow worse when using vector databases or recursive summaries to always fit the token maximum length prior to sending a request.

I hope my response isn’t too confusing, I didn’t really go down the rabbit hole of vector databases, so take this comment with a pitch of salt!

Thanks for letting me know @Thiago
I’m not familiar w/vector databases at all and I had no idea that using Pinecone would be an issue. Looking at the readme file it implied that pinecone allowed retention of the uploaded docs if the page was refreshed - so I just setup and account…

I’m looking for a way to query a laravel/react codebase in order to understand and document the code. ChatGPT works great, but uploading is tricky - even with plugins. Any suggestions would be appreciated.

Thanks again.