#meeting-1:fedoraproject.org log

<@marmijo:fedora.im>

15:30:49

!startmeeting fedora_coreos_meeting

<@meetbot:fedora.im>

15:30:51

Meeting started at 2025-09-24 15:30:49 UTC

<@meetbot:fedora.im>

15:30:51

The Meeting name is 'fedora_coreos_meeting'

<@marmijo:fedora.im>

15:30:55

!topic roll call

<@dustymabe:matrix.org>

15:33:06

!hi

<@zodbot:fedora.im>

15:33:08

Dusty Mabe (dustymabe) - he / him / his

<@siosm:matrix.org>

15:33:24

!hi

<@zodbot:fedora.im>

15:33:29

Timothée Ravier (siosm) - he / him / his

<@ydesouza:fedora.im>

15:33:29

!hi

<@zodbot:fedora.im>

15:33:31

Yasmin Valim de Souza (ydesouza)

<@hricky:fedora.im>

15:34:23

!hi

<@zodbot:fedora.im>

15:34:24

Hristo Marinov (hricky) - he / him / his

<@jbtrystram:matrix.org>

15:35:20

!hi

<@zodbot:fedora.im>

15:35:22

Jean-Baptiste Trystram (jbtrystram) - he / him / his

<@marmijo:fedora.im>

15:36:19

!topic Action items from last meeting

<@marmijo:fedora.im>

15:37:09

!info The normal community meeting was cancelled in favor of a live video meeting. no action items from the meeting on 2025-09-10.

<@aaradhak:matrix.org>

15:37:22

!hi aaradhak

<@zodbot:fedora.im>

15:37:24

Aashish Radhakrishnan (aaradhak)

<@mnguyen:fedora.im>

15:37:36

!hi mnguyen

<@zodbot:fedora.im>

15:37:38

Michael Nguyen (mnguyen)

<@dustymabe:matrix.org>

15:37:40

marmijo: we should probably discuss that briefly

<@dustymabe:matrix.org>

15:37:56

we should have probably tried to figure out some action items there, so maybe let's do it now

<@dustymabe:matrix.org>

15:39:52

5. start to investigate the leading option

<@dustymabe:matrix.org>

15:39:52

1. upload meeting recording somewhere

<@dustymabe:matrix.org>

15:39:52

2. reply to original invite with a link to the recording

<@dustymabe:matrix.org>

15:39:52

3. summarize discussion in https://github.com/coreos/fedora-coreos-tracker/issues/2006

<@dustymabe:matrix.org>

15:39:52

4. reach out to flatcar folks for input on what was discussed

<@dustymabe:matrix.org>

15:39:52

I'm thinking:

<@dustymabe:matrix.org>

15:39:52

<@dustymabe:matrix.org>

15:41:53

I have to go, brb

<@marmijo:fedora.im>

15:41:57

That sounds good to me. Does anyone want to take on some of these items?

<@marmijo:fedora.im>

15:43:15

Is there a good place to do this? Preferably somewhere in the Fedora ecosystem?

<@marmijo:fedora.im>

15:43:15

<@marmijo:fedora.im>

15:43:15

> upload meeting recording somewhere

<@ydesouza:fedora.im>

15:43:59

I can work on that, but I was wondering the same, where we could upload this recodingo :)

<@ydesouza:fedora.im>

15:44:08

I can work on that, but I was wondering the same, where we could upload this recording :)

<@dustymabe:matrix.org>

15:44:51

ok back

<@dustymabe:matrix.org>

15:45:11

the reason I bring up the 1-5 above is if we don't follow up, it's like the meeting never happened

<@marmijo:fedora.im>

15:45:48

+1. we need to follow up and start the investigation

<@dustymabe:matrix.org>

15:45:52

https://dustymabe.fedorapeople.org/meetings/

<@dustymabe:matrix.org>

15:45:52

<@dustymabe:matrix.org>

15:45:52

I've uploading video meetings before to my fedorapeople space, so we can just keep using that

<@dustymabe:matrix.org>

15:46:10

I'm guessing spresti has access to the meeting recording

<@dustymabe:matrix.org>

15:47:12

pinged him.. he's trying to connect

<@dustymabe:matrix.org>

15:47:18

let's see if he sorts it out

<@spresti:fedora.im>

15:48:11

!hi spresti

<@zodbot:fedora.im>

15:48:12

Steven Presti (spresti)

<@spresti:fedora.im>

15:48:48

Yeah I have access to the recording; I was trying to figure out how to make it public; Its currently locked in redhat access only.

<@dustymabe:matrix.org>

15:49:34

!action dustymabe will post the followup action item checklist to https://github.com/coreos/fedora-coreos-tracker/issues/2006 Yasmin will work with spresti on the action items. dustymabe will gain access to the recording and upload it to https://dustymabe.fedorapeople.org/meetings/

<@dustymabe:matrix.org>

15:49:43

sound good?

<@spresti:fedora.im>

15:50:03

Yep sgtm

<@marmijo:fedora.im>

15:50:11

<@ydesouza:fedora.im>

15:50:16

Thanks! +1

<@marmijo:fedora.im>

15:50:34

Thanks spresti and Yasmin Valim de Souza !

<@marmijo:fedora.im>

15:51:08

Let's move on to the next topic

<@marmijo:fedora.im>

15:51:16

!topic Review Fedora 43 Release Schedule

<@marmijo:fedora.im>

15:51:24

!link https://fedorapeople.org/groups/schedule/f-43/f-43-key-tasks.html

<@marmijo:fedora.im>

15:51:47

!info Fedora 43 CoreOS beta was released in the `next` stream with version: `43.20250917.1.0`

<@marmijo:fedora.im>

15:51:52

!link https://lists.fedoraproject.org/archives/list/coreos@lists.fedoraproject.org/message/GK4RMQ3UFMLJGKMBUVTTWGNFVFNNSH5E/

<@marmijo:fedora.im>

15:52:53

Next up in the list is "Retire FTI packages in NEW state" which I don't think will affect our process

<@marmijo:fedora.im>

15:53:14

After that, we're looking at Final Freeze starting on `2025-10-07`

<@dustymabe:matrix.org>

15:55:38

I think on us right now is our testing. We have a test day/week ongoing this week.

<@ydesouza:fedora.im>

15:57:16

Feel free to help to test this week! All info here: https://fedoraproject.org/wiki/Test_Day:Fedora_43_CoreOS

<@zodbot:fedora.im>

15:57:44

dustymabe has already given cookies to ydesouza during the F42 timeframe

<@marmijo:fedora.im>

15:57:49

!info Fedora 43 CoreOS beta Test Week is ongoing. See https://fedoraproject.org/wiki/Test_Day:Fedora_43_CoreOS for details and post test results here: https://testdays.fedoraproject.org/testday/6

<@zodbot:fedora.im>

15:58:04

marmijo has already given cookies to ydesouza during the F42 timeframe

<@zodbot:fedora.im>

15:58:34

No Fedora Accounts users have the @aaradhak:matrix.org Matrix Account defined

<@marmijo:fedora.im>

15:58:58

Alright, we have a few topics to cover today.

<@marmijo:fedora.im>

15:59:04

!topic Introducing the fcos-assistant Bot

<@marmijo:fedora.im>

15:59:11

!link https://github.com/coreos/fedora-coreos-tracker/issues/1999

<@ydesouza:fedora.im>

16:00:32

Bring this topic again to discuss some concerns about the AI bot.

<@spresti:fedora.im>

16:01:43

Regarding Timothée Ravier (travier)'s worries, I think the issue is that the bot might be seen as a source of truth? which using a model which is not directly trained on our docs might mean hallucinations. Im not sure what the correct approch would be currently prompting is the easy route, do we use somthing like Ilab to train our own llm using docs?

<@spresti:fedora.im>

16:02:21

The problem then becomes cost; 1. training and then 2. hosting said model

<@dustymabe:matrix.org>

16:02:53

spresti: yes, cost, but more so 3. engineer $time

<@dustymabe:matrix.org>

16:03:56

IoW if it's decently good without having to invest more, that would be nice. But either way I think we'd be monitoring the performance of the bot here to see how useful or useless it is, and then decide how we wanted to proceed.

<@dustymabe:matrix.org>

16:04:34

definitely interested in Timothée Ravier (travier) speak here, though

<@ydesouza:fedora.im>

16:04:42

We had some experience training a model with only butane docs and took a really long time. So yeah, its a lot of investment.

<@dustymabe:matrix.org>

16:05:37

semi-related: aaradhak and I did verify we were able to sign `coreosbot` FAS account into matrix and join a few channels. It's not wired up to any backend software at this point, but when we're ready we should be able to get a token and use that account.

<@ydesouza:fedora.im>

16:06:10

Yes, I am also worried about what Timothée Ravier (travier) comment that the Fedora infra has been having trouble with agressive AI scrappers/bots.

<@ydesouza:fedora.im>

16:06:51

I want to test it again to monitor the bot performance, but I need to talk with infra first :)

<@spresti:fedora.im>

16:08:02

Agreed, and we could maybe update the prompt to always indicate that it could be wrong and to do your own fact checking.

<@spresti:fedora.im>

16:08:02

<@spresti:fedora.im>

16:08:02

Yasmin Valim de Souza Yeahhhh and I think it would be worse; We would want to train many more iter's and with more data, but maybe its gotten better?

<@spresti:fedora.im>

16:08:02

Well AI goal is coming up (30th) should we try this now? Not sure of a possible negative that we cannot recover from?

<@dustymabe:matrix.org>

16:08:18

if that is a problem we can just point the bot at github (which is the source of truth for our docs)

<@dustymabe:matrix.org>

16:08:18

<@dustymabe:matrix.org>

16:08:18

> Fedora infra has been having trouble with agressive AI scrappers/bots.

<@dustymabe:matrix.org>

16:08:18

<@dustymabe:matrix.org>

16:08:30

<@dustymabe:matrix.org>

16:08:30

> Fedora infra has been having trouble with agressive AI scrappers/bots.

<@dustymabe:matrix.org>

16:08:30

if that is a problem we can just point the bot at the github repo (which is the source of truth for our docs)

<@dustymabe:matrix.org>

16:09:50

I mean, I'm open to still change with infra, but we have an alternative

<@dustymabe:matrix.org>

16:09:58

I mean, I'm open to still change with infra, but we have an alternative and shouldn't be blocked by it

<@ydesouza:fedora.im>

16:10:46

I guess it's a good idea and we can use github for that, no problem :)

<@dustymabe:matrix.org>

16:11:08

I can try to come up with a summary for us to past back to the ticket here if that would be useful?

<@ydesouza:fedora.im>

16:11:34

Sure!

<@ydesouza:fedora.im>

16:12:37

I will talk with you or Aashish Radhakrishnan to see how can I have access to the bot account and run a new test with a better prompt and having the github repo as source of true.

<@dustymabe:matrix.org>

16:13:22

!proposed Without training our own model we can't guarantee answers to questions will only come from our docs, but we can include in the instructions to the LLM to only consider our docs as input and monitor the quality of the results to determine if they are useful or not. Based on that we can determine if we need to invest more in tuning. As for scraping our Fedora docs site aggressively, an alternative is to point the tool directly at our docs github repo, rather than the docs website.

<@marmijo:fedora.im>

16:14:03

<@aaradhak:matrix.org>

16:14:07

<@ydesouza:fedora.im>

16:14:08

<@marmijo:fedora.im>

16:15:24

Any other votes for or against?

<@jbtrystram:matrix.org>

16:15:55

<@mnguyen:fedora.im>

16:16:02

<@hricky:fedora.im>

16:16:16

<@marmijo:fedora.im>

16:17:08

!agreed Without training our own model we can't guarantee answers to questions will only come from our docs, but we can include in the instructions to the LLM to only consider our docs as input and monitor the quality of the results to determine if they are useful or not. Based on that we can determine if we need to invest more in tuning. As for scraping our Fedora docs site aggressively, an alternative is to point the tool directly at our docs github repo, rather than the docs website.

<@marmijo:fedora.im>

16:17:30

🧑‍⚖️

<@marmijo:fedora.im>

16:18:20

Are there any other comments, or more to discuss here? We have time for another topic.

<@marmijo:fedora.im>

16:18:52

!topic tracker: Fedora 43 changes considerations

<@marmijo:fedora.im>

16:18:59

!link https://github.com/coreos/fedora-coreos-tracker/issues/1934

<@marmijo:fedora.im>

16:19:24

I ran the script and there were 3 new changes reported

<@marmijo:fedora.im>

16:20:12

change 238

<@marmijo:fedora.im>

16:20:21

!link https://fedoraproject.org/wiki/Changes/FoomaticRipRejectsUnknownValues

<@marmijo:fedora.im>

16:21:04

I think there's no impact to FCOS here

<@dustymabe:matrix.org>

16:21:05

printer stuff? I think we can skip

<@marmijo:fedora.im>

16:21:31

239

<@marmijo:fedora.im>

16:21:35

!link https://fedoraproject.org/wiki/Changes/Greenboot_RS_Change_Proposal

<@dustymabe:matrix.org>

16:22:17

We don't ship greenboot, but have wanted to in the past

<@marmijo:fedora.im>

16:22:43

okay, we can skip this one too

<@marmijo:fedora.im>

16:23:15

240

<@marmijo:fedora.im>

16:23:19

!link https://fedoraproject.org/wiki/Changes/NodejsAlternativesSystem

<@marmijo:fedora.im>

16:23:29

we can probably skip this too. we don't ship nodejs

<@marmijo:fedora.im>

16:23:52

And that's all for the new changes. Is there anything we need to discuss regarding the other changes in the list?

<@dustymabe:matrix.org>

16:24:14

<@dustymabe:matrix.org>

16:24:17

thanks marmijo

<@dustymabe:matrix.org>

16:24:33

pretty soon we'll be able to review F44 changes :)

<@marmijo:fedora.im>

16:25:33

yeah! the cycle continues!

<@marmijo:fedora.im>

16:27:09

Let's move to open floor since we only have a few minutes left

<@marmijo:fedora.im>

16:27:30

!topic Open Floor

<@dustymabe:matrix.org>

16:29:22

It's relatively minor, and we'll hopefully get it fixed before rolling it out to `testing` and `stable`.

<@dustymabe:matrix.org>

16:29:22

<@dustymabe:matrix.org>

16:29:22

There's one known issue from the recent container signing migration: https://github.com/coreos/fedora-coreos-tracker/issues/2035

<@dustymabe:matrix.org>

16:29:40

!info There's one known issue from the recent container signing migration: https://github.com/coreos/fedora-coreos-tracker/issues/2035

<@marmijo:fedora.im>

16:30:58

Thanks all for joining! Have a great day!

<@marmijo:fedora.im>

16:31:05

!endmeeting