A package is a folder of modules. Its __init__.py runs once, the first time anything from the package is imported, and the names it defines become the package's attributes. Technically a package is a module with a __path__: the list of folders its submodules are loaded from.

import importlib
import os
import sys
from pathlib import Path

Path("shop").mkdir(exist_ok=True)
Path("shop/prices.py").write_text(
    "TAX = 0.2\n")
Path("shop/__init__.py").write_text(
    "from .prices import TAX\n"
    "__all__ = ['TAX']\n")
sys.path.insert(0, os.getcwd())
importlib.invalidate_caches()

import shop
print(shop.TAX, shop.__all__)
print(shop.prices.__name__)
print(shop.__path__)

• from .prices import TAX is a relative import: . is this package and .. its parent. It works only in a module imported as part of a package.
• A submodule loads only when something imports it. Here import shop loads shop.prices because __init__.py imports it.
• __all__ lists the names that from shop import * binds. Without it, every name not starting with _ is bound.

Keep __init__.py short: re-export the public names and leave the real work in submodules.

Write is_package(module), true when the module has a __path__. Write public_names(module) returning what from module import * would bind: module.__all__ as a list when it exists, otherwise the sorted names not starting with _. Write submodule_names(package) returning the sorted names of every module in the package's folders, imported or not, using pkgutil.iter_modules.

The explanation opens here once every check passes.