kavacha

Install

pip install kavacha
pip install "kavacha[window]"
pip install "kavacha[macapp]"
pip install "kavacha[windows]"

The extras add native-window support and platform freezer dependencies.

Describe the app

from kavacha import App, tree, doc_types, mypyc_modules

app = App(
    name='Demo',
    entry='demo_app.py',
    version='1.2.3',
    icon='assets/Demo.icns',
    packages=['demo', 'fasthtml', 'uvicorn', 'numpy'],
    includes=['demo.cli', *mypyc_modules()],
    grafted=['apsw', 'playwright'],
    data=tree('demo/static', 'demo/static'),
    extras=['desktop'],
    doc_types=doc_types(['.py', '.rs', '.md']),
)
app.py2app_options()

One App specification supplies both platform freezers. Explicit packages and modules cover imports a freezer cannot discover.

Build

from kavacha import build, check

check(app, root)
build(app, root, setup_py='packaging/macos/setup.py')

build uses a compatible interpreter, installs pinned requirements from uv.lock, refuses to replace a running bundle unless forced, and records the source commit.

After freezing

from kavacha import finish

finish(bundle, app)

finish grafts archive-incompatible packages, removes unreachable archive copies, links duplicate files, mends Frameworks libraries two sources share a basename for, and installs the modern icon before signing.

Icons

from kavacha.icons import icons_for

icons_for('assets/logo.png', 'assets', 'Demo')

A square image of at least 512 pixels produces platform and web icon assets. Invalid sources are rejected rather than upscaled.

Window

from kavacha.window import run_shell, shell_ready

ok, why = shell_ready()
run_shell('http://127.0.0.1:8000', title='Demo', size='1440x900')

run_shell opens a pywebview window over a loopback server. The host supplies menus and workspace callbacks.

Status

The spec, build driver, bundle surgery and icons are covered by tests that run anywhere. The window and menus are ported working code, but the platform they matter on is not one CI can exercise here — treat that module as needing a real run on macOS before you rely on it.