PyWhales

The Docker and its composer client for Python, designed to be fun and intuitive!

Works on Linux, macOS and Windows, for Python 3.14.10 and above.

• Documentation
• GitHub repository

How to install?

uv pip install pywhales

See the Docker CLI guide for instructions on installing Docker.

Some cool examples

Start by doing

from pywhales import docker

and then:

• docker run hello-world -> docker.run("hello-world")
• docker pull ubuntu -> docker.pull("ubuntu")
• docker build ./ -> docker.build("./")
• docker compose up my_service -> docker.compose.up(["my_service"])
• docker image ls -> docker.image.list()
• docker ps -> docker.ps()
• docker cp -> docker.copy()

>>> from pywhales import docker

>>> output = docker.run("hello-world")
>>> print(output)

Hello from Docker!
This message shows that your installation appears to be working correctly.

...

>>> from pywhales import docker
>>> my_image = docker.build(".", tags="some_name")  # uses Buildx/buildkit by default
[+] Building 1.6s (17/17) FINISHED
...

Some more docker.run() advanced examples with postgres

docker run --name some-postgres -e POSTGRES_PASSWORD=mysecretpassword -d postgres

becomes

from pywhales import docker

docker.run(
    "postgres:9.6",
    name="some-postgres",
    envs={"POSTGRES_PASSWORD": "mysecretpassword"},
    detach=True,
)
print(docker.ps())
# [pywhales.Container(id='f5fb939c409d', name='some-postgres')]

docker run -it --rm --network some-network postgres psql -h some-postgres -U postgres

becomes

from pywhales import docker

# since it's interactive, you'll be dropped into the psql shell. The python code
# will continue only after you exit the shell.
docker.run(
    "postgres:9.6",
    ["psql", "-h", "some-postgres", "-U", "postgres"],
    networks=["some-network"],
    interactive=True,
    tty=True,
    remove=True,
)

docker run -d --name some-postgres -e POSTGRES_PASSWORD=mysecretpassword -e PGDATA=/var/lib/postgresql/data/pgdata -v /custom/mount:/var/lib/postgresql/data -v myvolume:/tmp/myvolume postgres -c shared_buffers=256MB -c max_connections=200

becomes

from pywhales import docker

docker.run(
    "postgres:9.6",
    ["-c", "shared_buffers=256MB", "-c", "max_connections=200"],
    name="some-postgres",
    envs={
        "POSTGRES_PASSWORD": "mysecretpassword",
        "PGDATA": "/var/lib/postgresql/data/pgdata",
    },
    volumes=[
        ("/custom/mount", "/var/lib/postgresql/data"),
        ("myvolume", "/tmp/myvolume"),
    ],
    detach=True,
)

Any Docker object can be used as a context manager to ensure it's removed even if an exception occurs:

from pywhales import docker

with docker.volume.create("random_name") as some_volume:
    docker.run(
        "postgres:9.6",
        ["-c", "shared_buffers=256MB", "-c", "max_connections=200"],
        name="some-postgres",
        envs={
            "POSTGRES_PASSWORD": "mysecretpassword",
            "PGDATA": "/var/lib/postgresql/data/pgdata",
        },
        volumes=[
            (some_volume, "/var/lib/postgresql/data"),
            ("myvolume", "/tmp/myvolume"),
        ],
        detach=True,
    )
    # so some stuff here

# here we are out of the context manager, so the volume has been removed, even if there was an exception.

Main features

• 1 to 1 mapping between the CLI interface and the Python API. No need to look in the docs what is the name of the function/argument you need.
• Support for the latest Docker features: Docker buildx/buildkit, docker run --gpu=all ...
• Support for Docker stack, services and Swarm (same API as the command line).
• Progress bars and progressive outputs when pulling, pushing, loading, building...
• Support for some other CLI commands that are not in Docker-py: docker cp, docker run --cpus ... and more.
• Nice SSH support for remote daemons.
• Docker object as Python objects: Container, Images, Volumes, Services... and their attributes are updated in real-time!
• Each Docker object can be used as a context manager. When getting out of the context, the Docker object is removed automatically, even if an exception occurs.
• A fully typed API (ty and IDE-friendly) compatible with pathlib and os.path
• All Docker objects and the Docker client are safe to use with multithreading and multiprocessing.
• Enable execution diagnostics by setting PYWHALES_DEBUG=1. Diagnostics omit command arguments, environment values, and captured output by default. Use DockerClient(raw_diagnostics=True) only when detailed diagnostics are appropriate for that client's logs. Only the exact value 1 enables debug output.

What about the license?

It's a MIT license, so quite permissive.

The license can be found in the git repository.