Hi again Oleg,
> > I think you may help! The identification of the group is still human
> > decision making process and I'm not sure it may be automated in any
point.
>
> I can certainly help with this then. I'll have some free time on
Friday and
> I can coordinate with you then.
Ah, I lost track of time..
Okay, so I wrote a grotesque shell script that extracted the name, synopsis,
and description from each package. Then I manually chunked the output into a
transformer, asking it to:
"Given a list of packages and descriptions, please[1] suggest ways of
categorizing them (by function or theme) in a way that keeps each
category
about equal size."
...followed by 5 paragraphs of minutia to make sure it does the right thing.
Then I give it 200 lines of packages ~36 packages. It gives me its output
which is inevitably missing one or more of the constraints I gave it, so I
say "Did you follow my instructions?" [2] After this, it either gives me an
output with one less thing wrong with it and an apology, or goes off in a
huff and wastes all my GPU credit.
Things that are standing out so far are:
- golang-web.scm could become a broader golang-networking.scm
- golang-crypto.scm could become a broader golang-security.scm
- There could be a need for a golang-text.scm that focusses on text
parsing and
processing.
- A golang-utils could exist, however this could easily become the 'other'
file.
Anyway, point is: I've got a very poor classification of packages in
golang.scm. I want to refine that using a langchain script. However, before
I move anymore forward with this, I wanted to ask: Is it okay to use AI
tools that are non-free in the aid of the GNU project?
> > We may extend handy script accelerating committing process, see
> > "etc/committer.scm"
>
> Okay, cool, I'll have a look at it on Friday.
Great, I haven't had a chance to look at this yet, but I will early next
week.
Here's a link to my repo: https://github.com/cdo256/guix-package-refactor
Let me know what you think.
Kind regards,
- Christina
[1] I am always polite to AI.
[2] Welcome to modern programming, people. Feels more like primary school
children.
