Requirements and dependencies¶

This chapter records the software Prodockit uses, the version floors declared by the package, and the external tools that Python package installation cannot supply. Authors normally do not need to select these versions individually: use prodockit pins whenever a project must be returned to the supported, tested combination.

The supported preparation route uses Python 3.14. The package's technical floor remains Python 3.10, and installation is tested on 3.10, 3.11, 3.12, 3.13 and 3.14; pip refuses older versions before import.

Installed with Prodockit¶

Everything in this section is pulled in automatically by pip install prodockit, except where noted.

Table 10.1 lists the runtime dependencies installed with Prodockit and explains why each one is needed.

1. Requirements installed with Prodockit

Requirement	Needed for
`Markdown` (>= 3.10.3)	every extension
`zensical` (>= 0.0.59)	Zensical integration and `prodockit.zensical_macros`
PyMdown Extensions (>= 11.0.2)	`prodockit.steps` and `prodockit.tree` are built directly on the PyMdown Blocks API; `prodockit.pdf` also preserves the output of PyMdown features
`beautifulsoup4` (>= 4.12)	`prodockit.pdf`
`click` (>= 8.0)	the `prodockit` command-line tool
`PyYAML` (>= 6.0)	`prodockit adopt` support for Zensical projects that retain a compatible `mkdocs.yml` or `mkdocs.yaml` configuration filename
`packaging` (>= 24.0)	comparing an adopted project's recorded Prodockit version floor with the installed release
`pypdf` (>= 4.0)	`prodockit.pdf`
`tomli` (>= 2.0)	reading a template manifest on Python 3.10, where `tomllib` does not exist yet
`tomlkit` (>= 0.13.2)	editing Adopt's TOML configuration and review ledger while preserving existing comments and formatting
`pathspec` (>= 0.12)	respecting `.gitignore` when collecting documentation sources outside a Git repository
`pymupdf` (>= 1.24)	only the back-of-book index - `pip install prodockit[index]`

The floors in Table 10.1 are declared in pyproject.toml, and a test keeps the table in step with them. They are compatibility floors, not a recommendation to assemble a toolchain from each minimum independently. Run prodockit pins to select the supported set.

Not installed by pip¶

The requirements in this section are not supplied by pip install prodockit, and they differ in kind.

Table 10.2 identifies the external tools that pip cannot install and the features that use them.

2. Not installed by pip

Requirement	Needed for
`weasyprint` (>= 69)	`prodockit.pdf`. A Python package, but not a dependency of prodockit - install it yourself. `prodockit.pdf` runs its command-line rather than importing it
`pandoc` (>= 3, builds pin 3.10.1)	`prodockit.pdf`, and `prodockit.bibliography` even without a PDF build. Genuinely not a Python package - there is nothing for `pip` to install
`mermaid-cli`, `mathjax-full` (Node >= 22)	only Mermaid diagrams and TeX maths in the PDF
Chrome or Chromium	only Mermaid diagrams - `mermaid-cli` renders them through a headless browser
A citation style (`.csl`)	only `prodockit.bibliography`. The standard style is fetched and validated by Bootstrap or Adopt; custom styles remain author-owned - see below

The citation style is a download rather than a Python package. Pandoc resolves harvard-cite-them-right.csl from the directory it runs in. Bootstrap fetches that standard style for a new template project, and Adopt now offers the same validated download when an existing project's configuration names it but the file is missing. Adopt retains its validated download in Prodockit's native download cache, so a later offline Adopt run can restore a known-good copy.

The style is deliberately not committed: it is third-party content with its own licence and release cadence. Adopt preserves an existing file and never guesses a source for a differently named custom style; obtain that chosen style yourself and place it at the path configured by csl_style. A manual installation can fetch the supported standard style with:

curl -fsSL -o harvard-cite-them-right.csl "https://www.zotero.org/styles/harvard-cite-them-right"

WeasyPrint is worth separating from Pandoc rather than filing both as external binaries: one is a pip install away and the other is not.

Pandoc is version-sensitive in a way that changes output rather than breaking the build: a major version below 3 renders code blocks as justified prose, and the builds pin an exact release because 3.x releases have disagreed about the same source. Bootstrap installs and verifies Pandoc 3.10.1 inside the project's virtual environment on every supported operating system. The system package can remain at another version; activate the project environment to use its pin. See Version pinning and drift.

See PDF generation for how prodockit.pdf locates these tools and Known limitations for why the Node tools are needed. A build with neither Mermaid diagrams nor maths needs neither of them, and no browser.