Compatibility

Check this page before you install Citry on a new server, CI image, or platform. It lists the Python versions, systems, and browsers Citry supports, and the one case where installing it needs a compiler.

Python versions

Citry supports Python 3.11 through 3.15:

Python 3.11
Python 3.12
Python 3.13
Python 3.14
Python 3.15

See the Python release status for which versions are still maintained upstream.

Operating systems

Citry's tests run on Linux and Windows with every supported Python version, and on macOS with the oldest and newest. Citry should run on any operating system that runs Python.

Prebuilt wheels

Most of Citry is plain Python. The one compiled part is citry-core, a Rust package that parses and compiles your component templates. A normal pip install downloads a ready-made citry-core wheel, so you do not need a compiler.

Wheels are published for:

Linux on x86-64, x86, ARM64, and ARMv7, for both glibc (manylinux) and musl (musllinux), plus s390x and ppc64le on glibc
Windows, 64-bit and 32-bit
macOS, on Intel and Apple Silicon

These wheels cover regular CPython 3.11 through 3.15. Free-threaded CPython 3.14 and 3.15 and PyPy 3.11 have wheels only on Linux; elsewhere pip builds citry-core from source, as described below. On a free-threaded build, Citry installs and works, but for now importing citry-core turns the GIL back on for the whole process (#433).

The exact list is in the files on PyPI.

Citry writes the same pages on every interpreter, apart from the builtins difference with the sandbox off described below. On standard CPython it renders simple = "vue" components faster than on PyPy, Pyodide, or free-threaded builds.

Differences on PyPy

On PyPy, citry check and the editor check less. PyPy has no _symtable module, which two of their analyses use. So the editor and citry check --types give no types to a template that uses a name its template_data() method assigns. And citry check does not read the module-level values that js_data() sends, so it misses a Secret or Path constant sent to the browser. citry check prints a note saying so, and the editor shows a warning when it starts. These checks do not change how a page renders, and Citry still refuses a Secret in js_data() while it renders. For the full check, run citry check on CPython. Issue #434 tracks giving PyPy the same answers.

If you turn off the expression sandbox with Citry(sandbox_expressions=False), template expressions on PyPy can also call any Python builtin. A missing variable that shares a builtin's name, such as {{ id }}, also renders on PyPy (as <built-in function id>) instead of failing. Compute such values in template_data() instead, as the security page shows in Builtins are missing. Then both interpreters give the same result.

Build from source

When PyPI has no wheel for your platform, pip downloads the source of citry-core and builds it on your machine. That build needs Rust 1.97 or newer. Install it with rustup, then run the install again. Nothing else in Citry needs a build step.

Browsers

Citry's browser code is tested in Chromium, Firefox, and WebKit.

Citry loads its own pinned copy of Vue for interactive components. Do not mount another Vue application over HTML that Citry rendered. To use Vue helpers such as ref in your own code, take them from Citry.vue, so they come from the same Vue copy as the page.