I’m building a file retrieval GPT and would like to import PDFs that when people search for something relevant it would produce the PDF, a link, and Ideally a visual. Is this possible?
It’s currently not possible to generate or create files using ChatGPT. That being said you could potentially generate VBA code that churns out documents. Or you could potentially hook it up to a Zapier that converts text to PDF?
You can create lot’s of files.
I can create pictures, Excelsheets, text documents, CSV’s files, etc…
One can even upload a picture, let Vision extract data from the image, render a graph from it and download the results as an Excelsheet.
Demo
Take this picture as an example. Can you extract the text, so I can read it as plain text?
Reboot your life the easy way
From exercise to diet and sleep – tiny tweaks that make a big difference
The Sunday Telegraph
Children forced back into masks in lessons
Can you create a version with that text, but styled as a HTML page.
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>The Sunday Telegraph - Front Page</title>
<style>
body {
font-family: Arial, sans-serif;
margin: 0;
padding: 0;
background-color: #f0f0f0;
}
.container {
width: 80%;
margin: 0 auto;
background-color: #fff;
padding: 20px;
box-shadow: 0 0 10px rgba(0,0,0,0.1);
}
.header {
text-align: center;
border-bottom: 2px solid #000;
}
.main-heading {
font-size: 50px;
margin: 0;
}
.sub-heading {
font-size: 24px;
font-weight: normal;
color: #555;
margin: 0;
}
.article {
border-top: 1px solid #ddd;
padding-top: 10px;
margin-top: 10px;
}
.headline {
font-size: 32px;
font-weight: bold;
}
.content {
font-size: 20px;
margin-top: 5px;
}
</style>
</head>
<body>
<div class="container">
<div class="header">
<p class="sub-heading">Reboot your life the easy way</p>
<p class="sub-heading">From exercise to diet and sleep – tiny tweaks that make a big difference</p>
<h1 class="main-heading">The Sunday Telegraph</h1>
</div>
<div class="article">
<h2 class="headline">Children forced back into masks in lessons</h2>
<p class="content">Zahawi insists lessons will be 'face to face' as he announces plans to keep schools open.</p>
</div>
<!-- Additional content can be added here -->
</div>
</body>
</html>
Can you create the HTML file for me, so I can download it?
(this is the file I was offered to download)
Awesome, and now in PDF format!
(offered me a PDF file)
It’s currently not possible to generate or create files using ChatGPT.
Not srictly true. Open a chat with code interpreter and you can use python to write to ChatGpt’s local temp filesystem, then get a link to that file.
interesting. I’d like to use it as a search function for our company knowledge base. I’d like to be able to ask it a question such as " are there any SPIFFs going on this month" and in our knowledge base we store PDF marketing docs (some of which are SPIFF docs). I’d like it to find and return relevant docs to me.
Do you have an approximation of the size of your company’s knowledge base in terms of the number of documents or megabytes required? It’s possible this would be doable, or you may need to find a different methodology and it could still probably be doable.
Zack Martin
Owner
The Lodestone Group
Office: 417-815-0133
Cell: 417-496-2081
zack@thelodestonegroup.com
www.thelodestonegroup.com
I’ve found you can also create PowerPoints, both of a set of images created through DALL-E or text generated in GPT4. It might make sense to ask for a Powerpoint, edit and then export to PDF?
That works.
I’ve been upload about 500mb of PDF files and was able to search them like you mentioned (also crosslinking data from PDF 1 and PDF 2).
But it’s not always close to the ideal answer, so you can’t trust it for 100%.
Love those gpt4-v hacks, I think you are doing great work with that model. I did a similar thing with my handwritten notes, messy handwriting. And it worked great - issue with a few words that were poorly written, but overall I think this functionality is how I want to digitize my journals.