Meeting started at 2026-06-04 16:31:23 UTC

The Meeting name is 'Fedora AI-ML SIG Meeting'

!hi

tflink: Tim Flink (tflink)

!hi

!hi

Carol Chen: Carol Chen (cybette) - she / her / hers

Diego Herrera: Diego Herrera (dherrera) - he / him / his

!hi

Justin Wheeler: Justin Wheeler (jflory7) - he / him / his

Finishing another meeting – I'll be full-focused here in-person.

!hi

Gordon Messmer: Gordon Messmer (gordonmessmer)

enough folks to get started ?

ok, let's get the party started

Tom Rix: this one's yours

keeping ollama updated. now we are up to 2.23.4

in rawhide, right?

reviewing some of the newer integrations. yes in rawhide.

poolside is a dude, it has eula.

looking at droid now.

we have openclaw already.

a dud? I assume that was a typo. if not, I don't understand what that means

a dud. sorry.

no worries, just making sure that I wasn't missing something :)

hermes, i got in python-fire. so bare minimum should be buildable now.

i haven't checked.

looked at ghostty , a better than your average tty i think mac uses it. it doesn't build.

the pi agent is sitting in package review. anyone have time for a review ?

link to bz?

clawdhub is also.

I thought I had time after the last meeting, and I was wrong. I'm sorry about that. I think I have time this cycle, too.

https://bugzilla.redhat.com/show_bug.cgi?id=2464801

there is a followon emacs agent-shell uses this.

https://bugzilla.redhat.com/show_bug.cgi?id=2477433

does not strictly need ollama of course.

that's just what i am using.

vim i documented in the wiki.

any other editors folks want to see llm-ed ?

the last time I tried doing fancy things with vim (vscode-style language servers), it frustrated me to no end. maybe I'll try again

It is cool, nonetheless

I'll probably try out the Vim stuff at some point, when Flock is not eating my brain

Tom Rix is one of them, AFAIK :)

Gotta scratch that itch!

TBH, I think it is more Fedora-y anyways to add this sort of integration with full FOSS editors we have long-provided in Fedora than trying to make stuff like VSCode work

I mean, come on Microsoft, what are you paying your engineers for anyways? `/s` 😛

that's why i ask about other editors. i am covering mine. it would be good to cover others. just i am not the best one to do that. emacs agent-shell i am using daily now.

any more updates on this other than a casual "emacs vs. vim" thing that will never be solved?

Sorry, no more off-topic from me, other than a kudos for the ongoing packaging work

I might have cycles in Q3 or Q4 to help out more with package reviews, but for now, I'm observing

ok. any help is appricated. the hold up is going to be package reviews. open source is generating a lot of ai thingies,, i guess using ai makes more thingies possible.

Fighting fire with fire? 😄

yes. that is why we need more ai in fedora.

so that's it for me. pi package would be one i am most interested in seeing in.

is it ok if we do the repo topic next? I have another meeting in 10 min and I think this shouldn't take long

yeah, we can do that. I was saving it for last to give you more time but now works just as well

please let me know if anyone has objections in creating the repo!

we'll get the intern to join the meeting sometime and do a demo

no objections from me

+1 from me

ok thanks! feel free to ping me (or Justin Wheeler ) if y'all have any questions about these projects!

In general, I am excited to see more stuff like this land under our umbrella

And FWIW, me, Carol Chen, and Dominik Kawka are working very closely with the intern to help keep a quality standard here too

The idea is that it will have nice local dev environment instructions on how to use it

that's all, thank you

sounds good

Justin Wheeler: this one is yours

I put this on the agenda, but I mostly wanted to raise awareness and see if anyone else wanted to take a look

I have a #TODO to look again at the new changes, but with PTO on Friday and Monday, and then Flock the week after, outlook is not good for a timely response from me

I'll get to it, but yeah, really just searching for what people think about this. It has potential to turn into a high-traffic repo as a library, versus a single skill

I'll help take a look next week after pydata london

jflory7 has already given cookies to cybette during the F44 timeframe

I have no objection to the rework but I also haven't looked at it very closely. I should do so

do we want to set a soft deadline for a decision on this so it doesn't languish in "pending" for a month?

18 June meeting seems reasonable but it could be merged before then if 1 or 2 people have time to take a look

anything else on this today?

✋️

I'm unclear what the hand emoji means in this context, I assume that it's to hold on for a sec

Just a heads up, 've been working on the onnxruntime library now that protobuf is landing on rawhide (suposedly it lands today)

yep, that's the next topic

See ya Carol Chen 👋

jflory7 gave a cookie to dherrera. They now have 33 cookies, 3 of which were obtained in the Fedora 44 release cycle

yeah! that :)

Is there a TL;DNR on onnxruntime?

I wanted to give a heads up here today cos there are many changes comming with this

is the protobuf update blocked on anything or is it just stabilizing/testing in COPR?

the protobuf blocked the posibility to update to newer versions

since newer versions of onnxruntime depended on more modern protobuf tooling that were hard to patch out

I've been working on this since last week, but now it's on a state that can be tested. I still need to land a new dependency on rawhide too, but I already have someone to review that

for now I'm testing things on copr://dherrera/onnx-devel

another thing, I'm considering removing the onnxruntime-rocm subpackage, since upstream changed the way that engines are distributed

but I still need to test if their implementation is stable

migraphx

vs rocm ?

both are still rocm.. but migraphx is the new thing.

migraphx is one thing, but this is on how things are packaged, now migraphx would be included with the main onnxruntime package

if things work correctly

so no subpackage would be needed

ah.. so onnxruntime package would still work for rocm ?

👍️

for me that's a priority

groovy. please poke me for testing it. piper-tts i believe is using onnxruntime now.

if I see that things don't work, I'll just deploy things as we always have

yeah! lot's of dependencies on this lib right now :D

that would be it from me 👍️

sooo many people luv meee it hurts

sounds good. thanks for the update and for the work

that's it for the agenda topics today

anything else that folks want to bring up?

/me wonders if [@gordonmessmer:fedora.im](https://matrix.to/#/@gordonmessmer:fedora.im) wanted to chat AI Developer Desktop since he is here

I think we chatted a bit about it last time but deferred some since he wasn't here

My memory is a Flock-sized fog

I thought there would be an update after the last council meeting but I don't think it was discussed.

planned/stated

If I remember right, we had spent some time thinking about how the AI/ML SIG could be a stakeholder, officially

It's funny how two weeks ago can feel like two months sometimes

yeah, I don't know about the status of the initiative per se, but work on individual items is going ahead. I have a list of people who were interested in various specifics.

should the SIG be a stakeholder? I suppose I should have read through the logs from the last meeting from when I was on vacation

[@gordonmessmer:fedora.im](https://matrix.to/#/@gordonmessmer:fedora.im) On the Council side, I know Aoife was working on an official reply with some thoughts on the Initiative process as a whole, but I think that will come later. We were going to chat with you but there's been a lot of other stuff going on. It's in mind though

i think at the least the sig should not be surprised with these initiatives and this meeting is a good place to discuss them _before_ they are public.

Let me see what I can unearth

There is a strong consensus that the SIG should be a venue for discussing new AI-related ideas before they are made public.

The group discussed the recent "Fedora AI Developer Desktop" proposal, noting that it was not presented well to key stakeholders.

AI/ML SIG's Role as a Stakeholder:

The SIG should be a formal stakeholder in the AI Developer Desktop work, and this will be discussed with the proposal author in a future meeting.

The SIG's documentation needs improvement. A plan to migrate relevant content from the Wiki to the official Fedora Docs site will be discussed at the next meeting.

Erm, I think the AI summary feels a bit aggro there, but I guess that's the basis of what we talked about last time

Something something, Docs

yeah, I found Carol Chen's post in discourse with the meeting minutes, summary etc.

Superb, I forgot about that too

Link?

@jflory7 & @cybette: Create a draft of a “Getting Involved” page for the AI/ML SIG section on the Fedora Docs site before the next meeting.

AI/ML SIG: Revisit the topic of separating content between the Fedora Docs site and the Wiki in the next meeting.

I guess this makes sense in retrospect, but since the SIG operates in public, the idea that it should be discussed in the SIG before "in public" is a little odd.

> There is a strong consensus that the SIG should be a venue for discussing new AI-related ideas before they are made public.

Ooohhhh yeah the action item I totally didn't follow up on yet 😅

I guess this makes sense in retrospect, but since the SIG operates in public, the idea that it should be discussed in the SIG before "in public" is a little odd.

> There is a strong consensus that the SIG should be a venue for discussing new AI-related ideas before they are made public.

This feels like an AI-ism, I think we just meant here as a starting place

We have no secret cigar smoke-filled backroom 😝

Putting that down again so it doesn't get lost, because it is important but Carol and I have both been super busy

And it is sort of related to all of this business about the AI Developer Desktop too

We don't yet have a great "front door" to the SIG yet

Fedora has a lot of discussion channels, including at least mailing lists, Matrix, and Discussions. I think it would be good if documented processes specified where conversations should start, because I'm not aware of any status of any of those channels as the correct place for a conversation to start.

that reminds me that I need to follow up on something that was mentioned in the first thread while I was away. will get to that after the meeting

there are mailing list people who are unhappy that the conversation didn't start on mailing lists, etc.

I think as an interim thing though, I don't think people would be opposed to you engaging and planning work here in tickets and getting feedback from us though

This is totally valid, and it feels more like a thing for the Council / project governance to address. That's not on you

Oh, those people will always exist no matter what you do 😝

Sincerely: I'm sorry that I didn't' start the conversation here. I apologize.

It's our fault for deciding it was ever a good idea to launch Discourse without an idea of what would happen to mailing lists

that's a long brewing issue, IMHO. the whole discourse vs. mailing list debate isn't new, unfortunately

(our meaning governance, again)

I was told that discourse was going to replace the mailing lists gradually. has that changed?

I don't think anyone was really peeved, I think the intent was to offer this as a home for you to collaborate and do the actual technical work you are mapping out

One Day™

Soon™

i would rather not ask an obvious . 'How does this NEW thing, fit into the OLD thing ?' in the most public place. that is part of talking with the stakeholder before, else lots of awkward questions.

I dislike the idea that any place is more public than any other

the place the conversation is happening is the right place to ask questions

The challenge is that some places get a lot more heat and light than others, which is unfortunately what happened to you going the route as an Initiative, which is also what you were specifically guided to do by those in leadership roles

I think this SIG could be a good place to shape the ideas and get feedback from people who are deeper in the space

yes, in retrospect, I agree

I'm not sure if there is a concrete action here or not, but I think the intent was to offer this space as a collaborative, less-visible place to shape ideas before inviting the whole world for feedback (which again, is not your fault, it is a fault of governance)

Like, maybe we could have a standing topic in the agenda for this work, maybe

Or a Forgejo milestone in our ticket tracker

I wasn't going to bring this up here but since we're going this route, fine. My criticism of that proposal was not that I thought the SIG should gatekeep anything related to AI but when a member/participant doesn't even mention it in matrix comes across as a pretty clear message of at least non-interest about input from that SIG. It's definitely not the council's responsibility to seek out existing entities for consultation, at a minimum

Just throwing out some ideas.

but the discussion has more or less died down, for now at least

I'm sorry that I didn't start the conversation here. I'm not disinterested in the SIG's input. I thought the SIG members would offer their input in the discussion thread. I see that I made a mistake, and I'm sorry about that.

I apologize, again.

especially when a core tenant of the proposal was around the community

I think this is a constructive point on how we can make the SIG a better place for stuff like this. That's why I think we talked about new "get involved" docs two weeks ago, because it isn't really obvious today

I am really keen to help with this kind of work, I just have to get through Flock to spend more cycles on it. I think we can get to a better place on all sides for this. I don't read bad intent from anyone in this situation

Just lots of confusion and process burden making things messy

it is what it is. I'm not trying to make this into a "you must eternally apologize for things" kind of bit

the discussion on discourse seemed to go to a point where nothing new was being said so it should remain at steady state for a while

the fact that "AI" means different things to different people and provokes emotional responses doesn't help

anyhow, we're drifting pretty far and we're over time unless we want to farther rehash this

anything else before we close out the meeting and move back to #ai-ml:fedoraproject.org ?

yeah, I have another call scheduled just now

Let's keep this conversation going!

I think we all want this to succeed

its good to close out with matrix channel is place to stuffff

Thanks for chairing today [@tflink:fedora.im](https://matrix.to/#/@tflink:fedora.im)

I'll see y'all next time 👋🏻

ok. thanks for coming everyone

next meeting will be in 2 weeks on June 18