AEO and ideas for discoverabilty of Fedora docs

Project Discussion
1

In a recent discussion about packaging and upstream URLs, i was informed that there was a better data source for some scripts i was writing (with heavy AI assistance) to get some quick proof-of-concept numbers to feel out the different potential paths:

Given some earlier comments in the thread along the lines of…

I realized that I had been somewhat misled by the AI tool.

Ok yes big woop AI misleads people all the time, so why does this matter?

I am a newcomer to Fedora Packaging. I got involved because I am interested in having metadata exist to link fedora packages to their upstream forges (see the other thread). Because I’m unfamiliar with the relevant Fedora procedures, and didn’t yet know which approach was going to work, I was using an AI tool to write fast, imperfect proofs of concept. This led to me getting given a solution that involved making probably a lot more network requests to Pagure than I needed (had i known about this tarball in advance).

I wanted to throw this observation out there because I’m curious whether an effort to somehow optimize the fedora docs for AI chatbots (a term ive heard is AEO[1] - the AI version of SEO) could be a useful way to steer what is potentially a growing number of newly-empowered contributors into better paths.

I suspect this is probably just a “but with AI” version of the same discoverability problem that many large organizations (such as my university) experience as the number of things happening at once within the org grows.

Just posting this as a curiosity-driven spin-off topic, but im curious to hear what people think.

  1. Here is the playlist/youtube channel where i first saw this term: https://www.youtube.com/watch?v=MLKgbeDeCxU&list=PLvJ_dXFSpd2v9cYyMFGutv0LqaC6Pio06 ↩︎

2

There is an intern working on RamaLama RAG with some Fedora stuff, you might want to find them and have a chat. I’ll try and remember where they hang out.

I dont really know that our docs are ready to be an authoritative source on all Fedora yet, and less sure that an AI will add value over a well designed menu.

3

I think well designed menus is probably more along the lines of what I was thinking anyway. I guess essentially the broader point I was getting at was improving the self-serve discoverability: I had to post a discussion before a proven packager told me that there was an easier way to gather and scrape all the rpm metadata.

I suspect it will probably involve many layers, like:

  • well organized menus for regular browsing
  • sitemaps/SEO for search engine traffic
  • strategic links to guides throughout fedora projects (maybe linking to a newcomers guide or something on dist git homepage)
  • maybe an AI RAG tool as a chatbots for guiding people with fedora specific questions (like me asking about the best place to submit my change proposal)

I think overall I threw this thread out there as a way to loosely document the small friction points I faced as a returning Fedora contributor that I fear might create additional friction for a first time contributor

4

Yes, all good ideas, we have recently put a lot of energy into getting Docs going again, with the help of a Council Initiative, would you like to join us in the Matrix channel #docs:fedoraproject.org and kick some ideas around there?