ciparity 0.1.0

Find drift between your pre-commit hooks and your GitHub Actions steps.

ciparity

Your pre-commit hooks and your CI are supposed to run the same checks. Over time they stop. Someone bumps ruff in .pre-commit-config.yaml and not in the workflow, someone adds mypy --strict to CI only, someone adds a hook that CI never runs. Then the branch is green locally and red on push, or worse, green everywhere while a check quietly stopped running.

ciparity reads .pre-commit-config.yaml and .github/workflows/*.yml and prints where they disagree. Static parsing only: no network, no Docker, no API keys, and it never runs your tools.

pip install ciparity
ciparity .

pre-commit hooks: 5   CI steps: 4

ruff    version differs: pinned to different versions
            pre-commit: 0.5.0
            ci:         0.6.2
mypy    arguments differ: different arguments
            pre-commit: -
            ci:         --strict
pytest  not in pre-commit: runs in CI but is not a pre-commit hook
            ci:         ci.yml:test
python  python differs: pre-commit pins a python version CI never sets up
            pre-commit: 3.11
            ci:         3.12

4 difference(s).

Exit code is 1 when there are differences, so it works as a check.

What it looks for

Check	Example
Tool in one side only	vulture is a hook, no workflow runs it
Version drift	hook rev: v0.5.0 against pip install ruff==0.6.2
Argument drift	--strict passed in CI but not in the hook
Python version	default_language_version: python3.11 while CI sets up only 3.12

Versions are read from hook rev:, from pip install tool==x, uv tool install, pipx install, npm i -g tool@x, and from the version: input of known actions such as astral-sh/ruff-action. Commands are found inside run: blocks, including behind uv run, uvx, poetry run, npx and python -m.

Usage

ciparity [path] [--json] [--ignore pytest,codespell] [--exit-zero]

As a pre-commit hook:

repos:
  - repo: https://github.com/Topicspot/ciparity
    rev: v0.1.0
    hooks:
      - id: ciparity

Deliberate non-goals and limits

• Only GitHub Actions in this version. GitLab CI and others are not parsed.
• Only tools it recognises are compared. File hygiene hooks like trailing-whitespace are ignored on purpose, nobody runs those in CI and reporting them would be noise.
• If a workflow runs pre-commit run --all-files, the two sides are parity by definition and "missing in CI" findings are suppressed.
• Composite actions and reusable workflows are not followed, so tools that only run inside them are invisible.
• Flags such as --fix or --all-files are treated as mode flags and never reported.

Alternatives

• pre-commit.ci runs your hooks as a service, which removes the drift instead of reporting it. If you can use it, use it.
• act runs workflows locally, so you can see what CI does. It answers a different question and is much heavier.
• zizmor and actionlint lint the workflow files themselves. They do not know your pre-commit config.

License

MIT

☕ Support the author

This project is free and MIT-licensed. If it saved you time, you can send a coffee.

USDT, Tron network (TRC-20) only:

TS9ywGeSyKQxiCszdKCHLR8DRAsnYCosNN

Другие языки / Other languages

• Українська: проєкт безкоштовний. Якщо він заощадив вам час — можна підтримати автора, USDT у мережі Tron (TRC-20), адреса вище.
• Русский: проект бесплатный. Если он сэкономил вам время, можно поддержать автора, USDT в сети Tron (TRC-20), адрес выше.