prek and CI¶

prek¶

# prek.toml: the mirror installs the wheel of the same version from PyPI
[[repos]]
repo = "https://github.com/collectiveai-team/botica-pre-commit"
rev = "v0.14.0"

[[repos.hooks]]
id = "botica"

[[repos.hooks]]
id = "botica-project"

[[repos.hooks]]
id = "botica-agents-md"

[[repos.hooks]]
id = "botica-rubric"

The hooks read the selection from the configuration, as botica lint and the editor do. botica-agents-md and botica-rubric run botica rules export --check for the block of AGENTS.md and for .abide/rubric.json, on every commit, since a new file changes the rubric. The mirror's hooks are verbose: prek hides the output of a passing hook, and a size warning does not fail it (distribution/botica-pre-commit). The recipe base installs these hooks from catalog/recipes/files/prek-base.toml, and writes the selection to botica.toml.

CI¶

A job of GitHub Actions runs the same selection on the lines that a pull request adds, and writes each finding as an annotation of the diff. --diff needs the history down to the merge base, and <version> is the version that prek.toml pins:

- uses: actions/checkout@v7
  with:
    fetch-depth: 0
    persist-credentials: false
- uses: astral-sh/setup-uv@v10.3.0
- name: botica
  env:
    BASE_REF: ${{ github.base_ref }}
  run: uvx botica@<version> lint --diff "origin/$BASE_REF" --output-format github

Without --diff, the job lints every file. The recipe ci installs two workflows that read text and no file: conventional-commits.yml runs botica lint --select conventional-commits --pr-title="$PR_TITLE" on the title of a pull request, and commit-policy.yml runs botica lint --select no-ai-coauthorship --commits "$range" on its commits.

abide¶

The workflow .github/workflows/abide.yml, which the recipe base installs, judges the diff of each pull request: abide check --json --diff, judged a second time when the first run has a finding or fails, then botica lint --judged with --fail-on warning, and a table of the findings (--output-format markdown) in the summary of the job. It reads the secret TYPESAFE_AI_API_KEY and the variable TYPESAFE_AI_BASE_URL of the repository; a pull request from a fork gets no secret, and the job fails. A pull request of Dependabot gets no secret of Actions either, and the job is skipped. Judgment rules says what abide judges.