pyoci 0.1.0

OCI-runtime spec represented as python objects.

pyoci

What: A library to define OCI Runtime Specification compliant container instances.

When: When you need to run or modify a container at the lowest level, without containerd or docker/podman.

Why: The full OCI runtime spec can be quite large to read and even trickier to implement. This library saves you all the json-wrangling and validation, without abstracting any features away.

How: Under the hood, everything here is a msgpack Struct. These structs were generated from the original json-schema with help of datamodel-code-generator and then manually refactored by me.

Pros:

• Full control over the container.
• Compatible with many runtimes.
• Very lightweight[^1].

Cons:

• Requires low-level knowledge of how a container is constructed.
• Doesn't support dealing with images in any way (for now), you'll need to provide an already existing container root.
• Isn't well tested with runtimes (for now).

This is a low-level library. If you want to simply run a container, without configuring all the inner workings, i'd suggest docker-py.

This library is runtime-agnostic, so it doesn't provide a way to actually run the container. You'll need to pass the definition to an appropriate runtime yourself.

Also, I want to say a huge thanks to koxudaxi and other contributors for the awesome code generator!

[^1]: Pyoci should be very fast and memory/resource efficent thanks to use of msgpack. The performance of actually running the container will depend on your provided runtime. However, since using a runtime directly involves much less abstraction than containerd/docker/etc, this approach should be more efficent overall.