This plugin doesn’t work. Where do I go to contact the author guys？Also, the avatar of the model using the plug-in is black. I asked if it was GPT-4, and it said it was GPT-3. Which model is using the plugin?Let’s set up a complaint rating and comment section in the store. The plugin can’t be used. I don’t know where to contact the author. It’s too difficult for me!
5 Likes
there is no upload button if you are looking for that.
You just need to provide the link to the PDF so it can download it directly
…and also make sure to enable the plugin in your search 
I’m using the loadPdf function and it’s still not working:
{
“pdf_url”: “https://example.com/mydocument.pdf”
}
The ChatWithPDF plugin avatar appears in place of the ChatGPT avatar, but it just returns the same old boilerplate “I apologize, it seems there was an error bla bla” that we’ve all seen too many times.
I’ve tried using various different PDF URLs, e.g.:
- https://digitalphysics.ru/pdf/Kaminskii_A_V/I_Am_a_Strange_Loop--Douglas_Hofstadter.pdf
- https://www.senate.gov/civics/resources/pdf/US_Constitution-Senate_Publication_103-21.pdf
All of these types fail too.
ChatGPT suggested uploading the PDF to Google Drive and then modifying the URL, and offered this instruction:
You need to replace /view?usp=sharing with /preview in the URL. So it will look like this: https://drive.google.com/file/d/FILE_ID/preview
This failed in the same way.
I’m not seeing any obvious link to any official documentation or even a website at all for whoever developed this plugin (nor any other available plugin, for that matter). Sure would be nice if OpenAI made it as easy to use these plugins as they have made it to install them…
1 Like
I am getting this error when trying to use the CHATpdf plugin
Error getting system message: {“message”:“Could not parse OpenAPI spec for plugin: [‘Got status 503 when fetching OpenAPI spec from https://chatwithpdf.sdan.io/openapi.yaml’]”}
I found a way to make GPT-4 read PDF.
Private PDF upload tutorial:
Upload the PDF to dropbox, and after creating a sharing link, replace www.dropbox.com in the link with dl.dropboxusercontent.com (and make sure the suffix is PDF) and send it to GPT-4 for it to be read with the askyourpdf plugin.
3 Likes
Yes, your advice works!
Thank you very much.
1 Like
Wah… this didn’t work for me 
None of these tips worked for me either; as for now I’d say that the PDF plugin is defective.
Update: Chatwithpdf still doesn’t work (for me), but another plugin called AskYourPDF does work.
Having said that, this “working” PDF plugin only seems to analyze short chunks of the PDF at any given time, and can’t really speak about the entire document as a whole (in terms of summarizing, etc.)
So I guess the holy grail of “read this 100-page document for me a write a cohesive 1500-word summary of the whole thing” is still not here yet.
4 Likes
You can use JiggyBase (https://jiggy.ai) to upload your documents (pdf, ppt, csv, xls, etc) as collections and chat your data. We have submitted JiggyBase as a plugin to the Plugin Store. Once it is approved, you can chat your doc (privately) in ChatGPT plugin. If you have dev access, contact us ([my email]) and I can give you a link to install.
Let me know if anyone has dev access is interested in testing our chat your data plugin.
Amy
2 Likes
Interesting concept and can imagine this being very useful, but this is a huge blow:
Is it possible for ChatGPT to read through all these documents incrementally before formally answering? For example, suppose I upload a 100-page pdf into the collection, and ask ChatGPT to summarise its contents; but before fully summarising it, it sends back a message such as “Finished reading and analysing pages 1-8, will now read 9-16” etc? In other words, it’s reading the contents through ‘different’ prompts, bypassing the limit and hence allowing it to then fully summarise the document?
use the askyourpdf plugin, new since today I believe, works excellent. Upload pdf and copy paste the code in GTP-UI: Ask Your PDF - Interactive PDF Conversations powered by ChatGPT
1 Like
Hi everyone! I made ChatWithPDF. Apologies for the downtime, we’ve had nearly 500k requests hitting a small server with PDFs ranging from research papers to entire books. I have spent the last week with little to no sleep constantly scaling up infra and updating how we load balance and load PDFs.
Google drive is now supported and its getting robust day by day. Please bear with me and let me know if you have any issues or problems going forward!
The best way to use it is to simply upload it to a service like tmpfiles/.org(not sponsored) where we can easily pull it, do embeddings search on it and give you the best result. Or any other simple links that are easy to parse for a bot like ChatWithPDF to access.
8 Likes
Hi Surya! Thanks for the hard work.
Google drive is now supported and its getting robust day by day.
I tried to use a public Gdrive link and even after some moderations in different link notations it still did not work. Can you explain which Google Drive link notation and shared link settings would work?
1 Like
Hi Surya! Thank you for your work on creating the plugin!
I have tried to upload the file to google drive and follow the steps which provided by the plugin. But it still shows ‘there was an issue with the URL provided.’
Could you provide the specific steps which work? I also tried the same way on the Onedrive, but it seems the same errors there.
Also, are there any limitations on how many pages we can upload for the pdf? The size of the file? Thank you!
1 Like
is there anyway we can upload our pdf file from our computer? Would be great as well to chat with multiple pdf’s simultaneously.
1 Like
Has anyone tried exploring how far you get with just the Browsing plugin?
I’ve heard some people are able to let the Browsing model read through their (online-hosted) PDFs and it was perfectly able to answer questions about it.
1 Like
Hi again everyone! The plugin is now back up, fully functional, and am hoping to keep a more consistent uptime. It can handle PDFs of any length but might take some time for it to embed/load into our vector db. I haven’t tested the limits just yet.
Sometimes it might give an error but please just retry again, sometimes the service times-out while it is still being embedded/stores in the vector db in the background.
For example I was writing en essay for school and gave it a 300 page book to read – it failed the first time but I tried again and it gave me quotes relating to my theme for my essay.
2 Likes