ChatPDF.com - Chat with any PDF using the new ChatGPT API

I wrote my own Python script to feed chunks of the book to ChatGPT API and ask questions about each chunk and collect the results. Costs a lot less and I can customize it as I want.

1 Like

That doesn’t work. I paste my API key and it says “Please input your apiKey!” and doesn’t accept it.

1 Like

Wish this actually worked. Would be so helpful. Hope it does in future. Similar to others. It says it doesn’t have any info on pages.

1 Like

The same is happening to me, It just won’t accept my api key.

After a hard refresh, GO to developer console and clear application data.

Would it be possible to be able to add more than 1 pdf document? I have multiple pdf documents that collectively have the full documentation of a program I’m using.

1 Like

Can I feed the ai with several pdf about a topic and ask questions about them? Some of the pdf will have references to the other complementary pdf documents. Is this possible?
Thanks

1 Like

Check out https://customgpt.ai/ - you can add 55+ document formats (including PDFs) and build chatbots with gigs of data. Use coupon code : ONEMONTHOFF if you need to try it out. Good luck !

1 Like

Hi Xathis! Love your tool so much.
Followed you on instagram, have some questions, would love to chat in DM!
:smiley:

Out of the blue question, What provider or software are you using to store embedding? I’m in hunt for a good and cost effective vector storage solution. Would be interesting to learn about how you handle this.

Hello. Greeting for you work. Is thi true? In the “9. CONTRIBUTION LICENSE” section, the issue of how the company handles uploaded files (Contributions) is addressed. According to this section, by uploading your Contributions (files) to any part of the Services, you automatically grant ChatPDF a non-exclusive, transferable, royalty-free, fully paid, and worldwide license to use, copy, reproduce, disclose, sell, resell, publish, transmit, and distribute such Contributions for any purpose, whether commercial, advertising, or otherwise. This also includes the creation of derivative works or the incorporation of Contributions into other works, and the granting and authorization of sublicenses. Distribution may occur in any media format and through any media channels.
Although it is not specifically mentioned whether ChatPDF stores the files in a repository, the description of the license suggests that the company might store and archive the uploaded files in their systems. Furthermore, they reserve the right to edit, re-categorize, or delete Contributions at any time and for any reason, without prior notice.

1 Like

Great project! Another interesting one is ClearAI . They have a few neat features but the response time is a bit longer. I can save notes and get video summaries all in one project. Then export the whole chat history with notes as a PDF. You can use my referral link to sign up: Register - Looks like you get about 5 queries before you need to subscribe. I have the $5 plan now and you get unlimited files and 35 queries.

Take a look at ClearAI. They have two options, one to scan text, the other to ask general question.

@xathis, thanks for building this!

In limited testing, I observed ideal responses in all but a few documents and use cases. The experience is simple, smart, and well-designed as a means for everyday people to build solutions based on focused topics. I’ve used it to convince a number of clients and colleagues that domain-specific models are going to make everything quite a bit better in the near future.

I know it’s early, but have a number of other ideas concerning business models that align with your concept. When you get to the business development topic for your underlying infrastructure, feel free to contact me - would love to share a few vi$ion$ I have for this. :wink:

1 Like

I tried uploading a screenplay (that I had already read and synopsized) and asked ChatPdf to provide a 2-page synopsis. The first time, it hallucinated a story that wasn’t there, complete with a “wealthy businessman” antagonist named “Adam Rodin” (in reality, Adam Rodin is the screenwriter’s agent, as noted on the title page of the script). The next two times, it told me it couldn’t do a synopsis because the screenplay contained only “dialogue and brief descriptions” – which is the nature of screenwriting. It also seemed as if it couldn’t access all the pages of the document, even though I uploaded them all. Is ChatPdf just not optimized to handle the screenplay format?

This shows great promise for a use case I have not seen handled by anyone yet, which is to help with board game rules.

I used it today and it’s not quite there yet. It provided an answer that was clearly wrong and not based on anything in the rule book. So it got this information from somewhere else.

Also, this particular game has rules across 3 different PDF’s. A rule book, play book, and player aid. I need to be able to upload multiple PDF’s for it to get the correct and full information from.

Totally agree - can you elaborate on the “domain specific models” idea please? (We are about to add support for this in our system https://customgpt.ai/ and I want to understand if its the same thing)

@jason.a.matzner If you’d like, I can try it in our system (no cost to you!) – would love to see what our system returns in this case. (Just DM the link to the PDF(s) if possible)

@cybernate DM me the 3 PDFs and I can give this a try (in our system). We had a similar customer recently. I’m curious to see how our system does with your game rules documents. (Alternatively, you can try it yourself - just ping me for a coupon code)

Imagine I have a 200-page document about prescription drugs. I have a community of domain experts, such as pharmacists and cost analysts, who need to query this corpus. I need to integrate this chat experience into existing systems so that the chatting occurs adjacent to other system features.

1 Like

We already support this out-of-the-box and added widget and API support this week. See an example of exactly this that my son built with about 1000 biomedical research documents from a researcher: LevinBot

How this was built:

  1. Upload about 1000 webpages and PDFs from the researcher to CustomGPT.
  2. Embed the chat widget into the website.

As of yesterday, this can be queried via API. So that can be used to integrate into your existing systems.

One caveat: This still runs on turbo. It will be upgraded to GPT-4 with full chat history around this weekend.

1 Like