How to Use uv for Publishing a Package to Private GitLab and Install from It

[the-korean-coder]

Hi,

I'm looking for guidance on how to:

• Publish a Python package to a private GitLab repo using uv.
• Install the package from the private GitLab repo in another project.
  Thanks!

[FishAlchemist]

Q1: Publish a Python package to a private GitLab repo using uv.

After I created the deploy tokens, setting UV_PUBLISH_URL、UV_PUBLISH_USERNAME and UV_PUBLISH_PASSWORD, then uv publish
https://docs.gitlab.com/ee/user/project/deploy_tokens/
https://docs.astral.sh/uv/guides/publish/#publishing-your-package

Q2: Install the package from the private GitLab repo in another project.

For Q2, you should be able to find the answer in the GitLab documentation.
https://docs.gitlab.com/ee/user/packages/pypi_repository/

As for Q1, you can refer to it as well since they are quite similar.

Note: Currently, UV does not support .pypirc. #7676

Uploading packages locally is done this way for me, but it seems unnecessary to apply for Deploy tokens when uploading packages from GitLab CI/CD.

[the-korean-coder]

Thanks, i was able to publish my packages.
Now, how do I add this package to an uv project using uv add? What would the exact syntax or command be for that?

[FishAlchemist]

Thanks, i was able to publish my packages. Now, how do I add this package to an uv project using uv add? What would the exact syntax or command be for that?

I haven't used any PyPI packages that require authentication. However, UV usually implements PEPs (Python Enhancement Proposals) and doesn't always introduce non-standard behaviors. Therefore, formats that pip can parse often only require minor modifications to work with UV.

• This is the documentation for the project series: uv init, uv add, uv remove, uv sync, and uv lock.
  https://docs.astral.sh/uv/guides/projects/
• This is the document about the incompatibility between UV and PIP.
  https://docs.astral.sh/uv/pip/compatibility/#pre-release-compatibility

[konstin]

I think this has been sufficiently answered.

[timotk]

I had the same issue (with publishing), and it still took me some time to figure it out. So I wrote a blog post about it.

Basically it boils down to the following .gitlab-ci.yml:

stages:  
  - publish  

variables:  
  PYTHON_VERSION: "3.13"  
  UV_VERSION: "0.5.5"
  UV_PUBLISH_URL: "${CI_API_V4_URL}/projects/${CI_PROJECT_ID}/packages/pypi"  
  UV_PUBLISH_USERNAME: "gitlab-ci-token"  
  UV_PUBLISH_PASSWORD: "${CI_JOB_TOKEN}"  

before_script:  
  - pip install uv==${UV_VERSION}

publish:  
  stage: publish  
  image: python:${PYTHON_VERSION}
  script:  
    # Set the version in pyproject.toml to the current git tag  
    - VERSION=$(git describe --exact-match --tags)  
    - uvx --from=toml-cli toml set --toml-path=pyproject.toml project.version $VERSION  
    # Build and publish the package  
    - uv build  
    - uv publish dist/*.whl  
  only:
    - tags

• Gitlab provides the current Gitlab instance url (${CI_API_V4_URL}) and a token (${CI_JOB_TOKEN}).
• uv picks up the environment variables and uses those as credentials. This is a little bit easier to read then putting them in the command itself.
• I had to apply this trick for versioning since uv does not do dynamic versioning (yet).

Hope this helps.