Skip to content

Latest commit

 

History

History
146 lines (95 loc) · 6.64 KB

RELEASE.md

File metadata and controls

146 lines (95 loc) · 6.64 KB
Raw

Release & Version Identifier

Marie is shipped from two package management systems, PyPi and Docker Hub. This article clarifies the release cycle and version identifier behind each system.

PyPi package versioning

We follow PEP-440, and a form of semantic versioning as explained above.

To install the latest final release:

pip install -U marie-ai

To install a particular final release:

pip install marie-ai==x.y.z

The term "final release" is relative to "developmental release" as described below.

Install Marie with Recommended Extensions

pip install -U marie-ai only installs the core dependencies of Marie.

The recommended way of installing Marie is pip install -U marie-ai

"standard" include extra dependencies that enables:

  • Executor Hub + Docker support
  • FastAPI + Websocket support (required when using Flow(protocol='http') or Flow(protocol='websocket'))
  • the best compression via LZ4 algorithm
  • the best async eventloop management via uvloop

Other extension tags such as [devel] can be found in extra-requirements.txt.

Do I need "[standard]"?

Depends on how you use and distribute Marie.

If you are using/distributing Marie as a microservice, you often only need to install the core dependencies via pip install marie-ai.

Developmental releases versioning

One every master-merging event, we create early releases directly from source control which do not conflict with later project releases. The version identifier for development release is x.y.z.devN, which adds a suffix .dev, followed by a non-negative integer value N, which is reset on every release.

To install the latest development release:

pip install --pre marie

Version epochs

Release cycle and versioning

Marie is developed continuously by the community and core team. Updates are grouped and released at regular intervals to align with software development best practices.

Marie follows a form of numbered versioning. The version number of the product is a three-part value x.y.z where x, y, and z are the major, minor, and patch components respectively.

  • Patch release (x.y.z -> x.y.(z+1)): Contain bug fixes, new features and breaking changes. Released weekly on a Wednesday morning CET.
  • Minor release (x.y.z -> x.(y+1).0): Contain bug fixes, new features and breaking changes. Released monthly on the first Wednesday of the month CET. This release is more QA tested and considered more stable than a patch release.
  • Major release (x.y.z -> (x+1).0.0): Are released based on the development cycle of the Marie . There is no set scheduled for when these will occur.

The following example shows how Marie is released from 0.9 to 0.9.2 according to the schema we defined above.

Event e After e, pip install marie-ai After e, pip install --pre marie-ai After e, master __init__.py
Release 0.9.0 0.9.0 0.9.1.dev0
Master merging 0.9.0 0.9.1.dev0 0.9.1.dev1
Master merging 0.9.0 0.9.1.dev1 0.9.1.dev2
Master merging 0.9.0 0.9.1.dev2 0.9.1.dev3
Release 0.9.1 0.9.1 0.9.2.dev0
Master merging 0.9.1 0.9.2.dev0 0.9.2.dev1

Docker image versioning

Our universal Docker image is ready-to-use on linux/amd64, linux/armv7+, linux/arm/v6, linux/arm64. The Docker image name always starts with marie-ai/marie followed by a tag composed of three parts:

marie-ai/marie:{version}{python_version}{extra}
  • {version}: The version of Marie. Possible values:
    • latest: the last release;
    • master: the master branch of marie-ai/marie repository;
    • x.y.z: the release of a particular version;
    • x.y: the alias to the last x.y.z patch release, i.e. x.y = x.y.max(z);
  • {python_version}: The Python version of the image. Possible values:
    • -py310 for Python 3.10;
    • -py311 for Python 3.11;
  • {extra}: the extra dependency installed along with Marie. Possible values:
    • : Marie is installed inside the image via pip install marie-ai;
    • -standard: Marie is installed inside the image via pip install marie. It includes all recommended dependencies;
    • -devel: Marie is installed inside the image via pip install "marie[devel]". It includes standard plus some extra dependencies;

Examples:

  • marie-ai/marie:0.9.6: the 0.9.6 release with Python 3.7 and the entrypoint of marie.
  • marie-ai/marie:latest: the latest release with Python 3.7 and the entrypoint of marie
  • marie-ai/marie:master: the master with Python 3.7 and the entrypoint of marie

Image alias and updates

Event Updated images Aliases
On Master Merge marie-ai/marie:master{python_version}{extra}
On x.y.z release marie-ai/marie:x.y.z{python_version}{extra} marie-ai/marie:latest{python_version}{extra}, marie-ai/marie:x.y{python_version}{extra}

Six images are built, i.e. taking the combination of:

  • {python_version} = ["-py37", "-py38", "-py39"]
  • {extra} = ["", "-devel", "-standard"]

Manual Release Entrypoint

Manual release entrypoint is designed for authorized core developers of Marie.

Trigger weekly release manually

One can release a hotfix immediately without waiting for the weekly release. Here are the steps:

  1. Go to here
  2. Click on "Run Workflow" dropdown button on the right and input the release token and the release reason.
  3. Click on "Run Workflow" green button.

Rebuild all Docker images for certain git tag manually

  1. Go to here
  2. Click on "Run Workflow" dropdown button on the right and input the release token and the release reason.
  3. Click on "Run Workflow" green button.

Rebuild all Docker images

Note, the manual rebuild on Docker images will NOT update :latest-* and :x.y-* aliases.