Problems /Bugs with Deep Research Agent Output

Support
1

Hi - new here would appreciate some help:

I made a copy of DRA and changed one thing : output to be PDF as opposed to HTML. Error comes up as ‘INVALID PDF STRUCTURE’ (No podcast was required.)

  1. I noticed that, off the bat, an error message: “The variable {{stringResearch}} is mentioned but does not exist”

  2. The URL (in my case example: https://app.mindstudio.ai/agents/deep-research-b4cce085/run/cf20067b-163c-40c7-859b-651a8e3bbcc3) - is not readable? I am trying to run a TL/DR agent on the content because I want a short version and a long version of the output. I tried using NotebookLM to scrape as well, and also cannot be scraped.

What I am trying to do is create a PDF output so that I can repurpose it for consumption ‘offline’ and also be able to input the PDF as a ‘knowledge base’ for a chatbot.

TIA

1 Like
2

Hi, I think I have a hunch as to what is happening in both cases.

PDF Output
To get the output of the Deep Researcher as a PDF, you need to change both the Generate Asset block and the Display Content block (the final two blocks preceding the green end block in the main workflow). It sounds like you might have changed the Display Type on the final block, but the asset being generated is still an HTML file and not a PDF document, which would be the cause of that error. In the Output section of the Generate Asset block configuration, change the format from HTML to PDF (you can also configure the page size, etc. here).

The Generate Asset block creates an HTML document, PDF, or Image and saves the URL of that asset to a variable. The Display Content block, when configured using the Display Type option, simply renders contents of a URL (in this case, the URL of the asset generated by the preceding block).

Accessing the Output in Other Tools
There are likely two things happening here. The first is that the /run URLs are private and only accessible from your account (try opening in Incognito to see). You need to click “Share” and then make the asset public or unlisted in order for other tools like NotebookLM to be able to see it and scrape it.

The second snag here is that for security purposes, the content of an asset is rendered inside an iframe. So oftentimes tools and scrapers (including the MindStudio chrome extension) only access the top-level frame and cannot “see into” the nested content on the page. We are working on a fix for our own extension that should be out soon, but this issue will likely still affect other tools.

PS—it might be easier to just click “Print” from inside the Deep Research report output and choose “Save as PDF” that way!

Hope this helps!

3 Likes
3

Really appreciate the detailed response :slight_smile :grinning_face: — (I did try print, but it only does the page I am on.) I will play some more. Thx again.

4

Hi - and just thought I’d report back. Managed to get a PDF created and ‘displayed’ albeit with horrible line breaks — any tips for prompting for a better ‘format’?