Loading
Preparing the page
Public context is loading before any interactive tool code.
Preparing the page
Public context is loading before any interactive tool code.
Inspect changed files and added or removed lines, then create a conservative commit-message suggestion without uploading the diff.
Your checked result will appear here.
The parser follows file headers and changed-line markers. Those observable facts pass into a short draft; unknown intent stays unknown.
The method is deterministic. The same diff produces the same result, and no model fills gaps with a plausible story.
Paths come from diff --git and +++ b/ lines, not from prose near the code.
Added and removed content lines are counted while the diff's own header markers are excluded.
Documentation, tests and configuration receive recognizable types. New source-only files may receive feat; other source edits default to chore.
The result uses type(scope): subject. It does not invent a reason, ticket number or assurance that tests passed.
This fixture is covered by the product test. One added line and the documentation path support the type and scope; they do not reveal why the change was requested.
diff --git a/docs/release-notes.md b/docs/release-notes.md --- a/docs/release-notes.md +++ b/docs/release-notes.md @@ -1 +1,2 @@ Release notes +Document the local export boundary.
docs(docs): update release notes
Observed: 1 file · 1 added · 0 removed
Updated the release notesA prose description has no file header or changed-line evidence, so it is not treated as a diff.
A syntactically tidy message can still misrepresent the work. Replace the generic subject when you know the actual intent.
Decide whether the change is truly a feature, fix, refactor, test, documentation update or maintenance task.
Use the subsystem understood by your team; a directory name is only a fallback signal.
State the outcome or reason when the diff alone cannot express it accurately.
Add breaking-change notes, issue references and test details only when you have verified them.
The generator can summarize structure. It cannot know the product requirement, deployment risk, review history or whether the edited code works.