juniper-ci-tools 0.3.0

Shared CI / build tooling for the Juniper ML platform (dependency-documentation generator + workflow-script-path lint + AGENTS.md version-drift lint)

juniper-ci-tools

Shared CI / build tooling for the Juniper ML platform.

This package is the single source of truth for the dependency-documentation generator that historically lived as scripts/generate_dep_docs.sh (or util/generate_dep_docs.sh) in every Juniper repo. Three variants of that script drifted on origin/main; the 2026-05-20 bug fix in juniper-cascor#276 (switching the conda-dependency extraction from sed to awk to avoid emitting an invalid trailing prefix: / variables: key) shipped in 1 of 8 repos. This package distributes that fix as pip install-able tooling so it cannot drift again.

This work mirrors the juniper-doc-tools PyPI migration plan that addressed the analogous 2026-05-18 doc-link validator incident.

Installation

pip install juniper-ci-tools

This installs two console scripts:

• juniper-generate-dep-docs — dependency-documentation generator (the consolidated scripts/generate_dep_docs.sh port; see "Usage" below).
• juniper-lint-workflow-paths — lints that every python|bash <path> invocation in .github/workflows/*.yml references a file that exists on disk. Added in 0.2.0; consolidates the 6 byte-identical copies of util/test_workflow_script_paths.py that existed across consumer repos.

The package requires Python 3.11 or newer and depends on PyYAML.

Usage

Run from the root of any Juniper repo that has a pyproject.toml:

juniper-generate-dep-docs

Equivalent module form:

python -m juniper_ci_tools

This will:

1. Create conf/ if needed.
2. Back up any existing conf/requirements_ci.txt with a timestamp infix.
3. Render notes/PIP_DEPENDENCY_FILE_HEADER.md (substituting placeholders like <X.Y.Z ...>, <YYYY-MM-dd ...>, <Python Version>, <Pip Version>) and append pip list --format=freeze output.
4. If conda is on PATH: back up any existing conf/conda_environment_ci.yaml, render notes/CONDA_DEPENDENCY_FILE_HEADER.md, append the dependency block extracted from conda env export --no-builds (using the awk-equivalent logic that produces valid YAML), and validate the result with yaml.safe_load.

If conda is not available the conda step is skipped with a warning (matches the legacy bash script). If the generated YAML fails to parse, the command exits non-zero.

CLI options

Flag	Default	Purpose
--repo-root	cwd	Repo root containing pyproject.toml
--conf-dir	conf	Output directory
--notes-dir	notes	Directory containing header templates
--pip-header	PIP_DEPENDENCY_FILE_HEADER.md	Pip header template filename
--conda-header	CONDA_DEPENDENCY_FILE_HEADER.md	Conda header template filename
--pip-filename	requirements_ci.txt	Pip output filename
--conda-filename	conda_environment_ci.yaml	Conda output filename
--no-conda	off	Skip conda generation even if conda is installed
--no-yaml-validation	off	Skip yaml.safe_load on generated conda file

Library API

from juniper_ci_tools import generate_dep_docs

result = generate_dep_docs(repo_root="/path/to/repo")
print(result.pip_file, result.conda_file, result.yaml_validated)

See the package's juniper_ci_tools/generate_dep_docs.py docstring for the full surface.

Development

cd juniper-ci-tools
pip install -e ".[test]"
pytest

License

MIT. Copyright (c) 2024-2026 Paul Calnon.