8000 AI usage memo · Issue #1624 · winpython/winpython · GitHub
[go: up one dir, main page]
Skip to content

AI usage memo #1624

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
stonebig opened this issue May 30, 2025 · 3 comments
Open

AI usage memo #1624

stonebig opened this issue May 30, 2025 · 3 comments

Comments

@stonebig
Copy link
Contributor
stonebig commented May 30, 2025 •
edited
Loading

recipes, from time-consuming experience

errors:

  • prompting too simple: the AI doesn't respect any rule and can change things you don't want at first (like variable names)
  • not using multi-AI loost the comparison and compensation (like orthographic errors, what is generated different is what to look)
  • an easy before/after comparison test is usefull, better when they can auto-check themself, simpler if they write auto-test
  • not adding safe exit to llm , "do you think it's possible ? is there enough information to", may reduce hallucinations
  • accepting torents of lines of code... as AI can bloat you also on this
@stonebig stonebig pinned this issue May 30, 2025
@stonebig
Copy link
Contributor Author

diff.py:
let's try to use AI in a smarter way:
- run current and proposals from openai, copilote, mistralai, gemnicodeworks
- compare results (accept slight massage at interface naming, output must be same as original)
- compiles without hurting much interface: openai,, gemini
- don't crash and result ok: openai,
- crash and burn : gemini, mistralai
- change interface (remove a parameter): copilote
- lines_of_code (ignoring commente of gemini): original: 170 , openai: 129, copilote: 98, mistralai: 161, gemini: 350

in the end, still 1 day to:
- digest the AI change
- then do the re-engineering to move diff.py down

@stonebig
Copy link
Contributor Author
stonebig commented May 31, 2025 •
edited
Loading

package_index_markdown (try on copilote, as move between the project files)

i would like to have the core part of the "package_index_markdown" function pushed from make.py down to wppm.py. 
Goal is to have less things in make.py, and more in wppm.py so most of the feature is available from wppm

proposed the stupid/basic solution, good here

@stonebig
Copy link
Contributor Author
stonebig commented Jun 1, 2025 •
edited
Loading

adding safe exit to llm , "do you think it's possible ? is there enough information to", may reduce allucinations

to copilote:
could we make piptree working either on current installeed packages, either in a directory of packages ? it seemes all we need is reading the metadata from the archive like wheelhouse.py does

gives a general structure that works, but prefer doing integration by hand so far

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@stonebig
0