This topic is quite old but, I needed an offline version of the help document. As @Deepakpais mentioned, if you’re offline and click the Flame Help from within Flame, the offline html does say ”A local version of flame help can be installed from _______” which takes you to the above link.
It would be great if it was actually possible to have an offline version of the document. I am working on a tutorial series and wanted to feed it to notebookLM as a source but, no dice. Sooo I vibe-coded some scripts and got the html files into a pdf. Internal links are not working and I didn’t scrape the images.
It is enough for my purpose at the moment. But let me know if you need it so I can put some more time into it and try to get a better document, or let me know if you want it as is.
Here is the mishmash of a pdf that I scraped off of the online help pages. It doesn’t look good but it is a usable offline documentation. I just used it to check if it answered @Jonhollis’s question about deleting the markers and it answered.
Just upload it into notebookLM, wait for it to read through and ask any questions you might have.
PDF get out of date very quickly, this is why we do not provide our documentation as PDF.
So, if you want a local version of the Online help of the week, there are many applications that allow you to save a section of a web site as PDF. Google is your friend
What app did you use to do it? I’m trying to download the online documentation and I’d like to know how do it. Tipical apps like httrack don’t convert the scrapped website to a pdf (as far as I know)
Hey @kily
I actually used LLMs to create the scripts for me. Google’s Antigravity took it too seriously and tried to write a whole framework for it. Then it was a tennis match between chatGPT and Antigravity. I fed back the error messages so they get what the problems are. When one got stuck with one, I gave the code to the other.
Also had to install some libraries with Homebrew on the mac and used them to get it working. I was so sloppy doing it, I might need to do it from scratch when 2027 is released.
Hi again. @Sinan. I’ve been driving ChatGPT crazy for days now. After 342534 tests, It’s completely unable to detect the right hierarchy throught sections. The online help is frustratingly resistant to scraping. I also tried with Gemini, with even worse results. I gave it your PDF, and it told me it’s impossible to generate it without human intervention, based on the information that can be extracted from the page. Just one question. Did you edit manually the pdf?
Short answer: No, I did not manually edit the pdf.
But, it was a multi-step process. First I scraped the online docs and wrote them to my local folder as html files (1880 of them). Next I gathered them into one html file and then, converted that into a pdf.
I had to install some libraries with homebrew on my mac. Somewhere along the way I tried to do it in a docker container but couldn’t manage to be that tidy.
I can share my entire chat log with chatGPT if you would like to take a look. DM me for the link.
Yeah sure. I’m doing same. First scrap and then the pdf. But it’s impossible for chatgpt find the most basic hierarchy. It get lost after the second level. And it says is impossible figure it out taking the info of the dinamic html and blah blah blah. I’ll DM you. Thanks Sinan.
If you let an LLM use something like Playwright, that should enable powers of navigation to help with hyperlinks, such as on the Flame site. Although I’d never want to recommend anything that could be a ToS or copyright issue…
One thing that might help with the whole “offline help” demand would be… Autodesk maintaining official MCP servers for Flame documentation. This way we can have a version-aware LLM Flame knowledge base tool, and Autodesk can avoid the otherwise relentless scraping that might come from the user base. Reasonable feature request?
Just for kicks, here are MCP Generation tools that your LLM can build and use itself (bonus points for switching on YOLO mode and walking away):
I did install and use Playwright at some point. I’ll look into this stack. I was just planning how I should go at vibe-writing shaders in a more structured manner. I need to document and do version control.
And I totally agree ADSK should provide a way for LLMs to query the docs. Stephane did say PDFs get out of date quickly, but there must be some way out of this, for users’ sake.
