Set Up Deployment of ACCESS-Vis Package to ACCESS-NRI Conda Channel

[rbeucher]

Set Up Deployment of ACCESS-Vis Package to ACCESS-NRI Conda Channel

Description:

We need to automate the deployment of the ACCESS-Vis package as a conda package to the ACCESS-NRI conda channel. The deployment process should be managed using GitHub Actions to ensure automatic updates whenever a new release is created.

Action Required:

1. Set Up GitHub Action for Conda Deployment

   • Create a GitHub Actions workflow that triggers deployment to the ACCESS-NRI channel upon pushing a new version tag or release.
   • Ensure the workflow includes steps to:
     • Build the conda package using conda-build.
     • Upload the package to the ACCESS-NRI channel using anaconda-client.

2. Conda Configuration

   • Ensure that the meta.yaml file is correctly configured for the package, including dependencies, versioning, and other necessary metadata.

3. Conda Credentials

   • Store the Anaconda credentials (username and API token) as GitHub repository secrets (e.g., ANACONDA_USERNAME and ANACONDA_API_TOKEN).
   • Configure the workflow to securely use these secrets during deployment.

4. Versioning Strategy

   • Ensure that the workflow increments the version number correctly based on Git tags or releases in GitHub.
   • Provide instructions in the README.md or CONTRIBUTING.md for maintaining and tagging new versions.

5. Testing the Workflow

   • Test the GitHub Actions workflow with a test deployment to a separate conda channel or staging environment to ensure everything functions properly before live deployment.

6. Documentation

   • Update the README.md with installation instructions for the conda package from the ACCESS-NRI channel and mention the automated deployment process in the repository documentation.

Additional Resources:

• Information on how to set up the deployment can be found in the access-med-utils repository.
• Organisation recommendations and guidelines can be found here:
  ACCESS-NRI Documentation Guidelines.

[rbeucher]

I think we need a Lavavu conda package...