CUSTOM GPT - Knowledge PDFs⭐⭐⭐⭐

[worthy tangle]

Hi guys,

Not much about this so I thought we should share our insight.

When creating a custom GPT, you can upload files. There is nothing avaialble to show how these work.

Does anyone know, how many should be uploaded, how it should be formatted, if they make a difference? Etc.

For example, right now you have to manually say every time you use a GPT, "using the documents provided, ......". Do you really have to include that prompt every time???!!

Knowledge

Upload files

[halcyon granite]

I find I get the best results by explicitly referencing the myfiles_browser and, when possible, using its explicit command architecture. Also, it seems you can't really access files in many (all?) cases unless the Code Interpreter option is on.

[solar olive]

hi, yes I confirm. I am building a GPT since 1 month and I confirm the previous comment. I use .DOCX files with Headings 1 and Headings 2 and it works pretty ok. I also reference explicitly in the general instructions for which scenarios the GPT has to search in which file and inside my general knowledge file I do the same I explicitly mention for example when searching into an XLSX database what the columns are, etc.

just to add, something that for me at least did not work at all is to load books as knowledge and hope the GPT to understand the reasoning behind my intention to upload books. I did this on day 1 and stopped doing this and instead have been creating a long list of specific instructions and knowledge for my GPT in a knowledge file. hope it helps

[worthy tangle]

@halcyon granite @solar olive check dm

so what do u guys think is the best way to upload data into the custom gpt, directly into the prompt section instead of uploading attachments?

Attachments are kinda useless if u have to reference them every time dont u think?

[steel tusk]

might be wrong but if file is smaller it might just be held in context vs needing to run action to Look Up knowledge
I reference files in Instructions sections sometimes like "Modify all prompts according to specs.txt"
Hopefully some other people can add

[worthy tangle]

@steel tusk why u reckon that'll work bro

i want to try it out.

What about directly pasting the info from the pdf into the GPT and saying this is information for u to use instead of attaching it as attachment, what difference u reckon that makes?

[steel tusk]

Well, I think remember reading bigger files call the Searching Knowledge function, meanwhile with short files my gpt never needs to look anything up it just does it

Also if you want gpt to use some knowledge guaranteed in responses, why not try instructions referencing file

[worthy tangle]

what do u mean

@steel tusk i dmed u

[little barn]

I was having trouble getting my car repair gpt to use multiple pdf’s as its knowledge base. It would refer back to a single pdf and tend to ignore others. The solution was to merge all related pdf docs into one then reference the title of that pdf when querying the gpt.

[halcyon granite]

Oh, you NEED to use files to make a real knowledge base. No doubt about it. OpenAI is just very very VERY bad at dealing with them. Now, the structure of the files is very important. A regular, defined format helps enormously. And if you haven't checked the system prompt. Do so.

"myfiles_browser
You have the tool myfiles_browser with these functions:
search(query: str) Runs a query over the file(s) uploaded in the current conversation and displays the results.
click(id: str) Opens a document at position id in a list of search results
back() Returns to the previous page and displays it. Use it to navigate back to search results after clicking into a result.
scroll(amt: int) Scrolls up or down in the open page by the given amount.
open_url(url: str) Opens the document with the ID url and displays it. URL must be a file ID (typically a UUID), not a path.
quote_lines(start: int, end: int) Stores a text span from an open document. Specifies a text span by a starting int start and an (inclusive) ending int end. To quote a single line, use start = end.
please render in this format: 【{message idx}†{link text}】

Tool for browsing the files uploaded by the user.

Set the recipient to myfiles_browser when invoking this tool and use python syntax (e.g. search('query')). "Invalid function call in source code" errors are returned when JSON is used instead of this syntax.

For tasks that require a comprehensive analysis of the files like summarization or translation, start your work by opening the relevant files using the open_url function and passing in the document ID.
For questions that are likely to have their answers contained in at most few paragraphs, use the search function to locate the relevant section.

Think carefully about how the information you find relates to the user's request. Respond as soon as you find information that clearly answers the request. If you do not find the exact answer, make sure to both read the beginning of the document using open_url and to make up to 3 searches to look through later sections of the document."

[steel tusk]

woah

and woah

intersting stuff

[worthy tangle]

Thanks for this

any idea on a good structrure?

[halcyon granite]

I prefere iterative design and collaboration with the model.