Repository navigation
Add Kimün - #1916
Merged
Merged
Add Kimün#1916
Conversation
Contributing criteria check[PASS]
|
| Criterion | Result |
|---|---|
| Stars (min 20) | pass 25 stars |
| Human contributors (min 2) | pass 6 human contributors |
| Age (min 6 months) | pass created 247 days ago (at least 6 months) |
All tool eligibility criteria passed. Thank you for your contribution.
Member
|
Thanks |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds Kimün, a command-line tool for code health metrics written in Rust: a grade for the project, cyclomatic and cognitive complexity per function, duplication, code smells, and from git history hotspots, ownership, temporal coupling and the reach of a diff.
I am the author of the tool.
Tool requirements
README.mddirectly.AI tags, if applicable
On the
uses-llmtag: I left it out. Every metric Kimün reports is computed locally and deterministically; no model takes part in the analysis. Thekm aicommand is separate and optional: two of its subcommands only set up a coding agent to callkm(a skill and permission rules), and the third,km ai analyze, lets a model ask for those same metrics and write a narrative report from them, and needs an API key to run at all. Happy to add the tag if you read the rule differently.The language tags are the ones Kimün measures complexity for, not the longer list it only counts lines of.