prompt_processing

Prompt Processing workflows

Prompt Processing is a Rubin Data Management project, and follows the DM Development Workflow. This guide summarizes team-specific details needed to support our independent release process or deployment management. It assumes familiarity with the application structure described in Core Concepts.

Jira

On Jira, we mark Prompt Processing related work items in several ways:

GitHub pull requests

Review and Testing

Prompt Processing pull requests should be reviewed per the usual procedure. Because Prompt Processing is not part of Science Pipelines and is not a library for any other package, it need not be tested with the stack-os-matrix Jenkins job (though it can be tested there by explicitly listing the package).

Instead, you should run integration tests of the service as described in the Playbook. GitHub actions automatically build service and initializer images whenever the pull request is updated. Even if you’re a long way from having something fit for review, it can be useful to create a draft pull request to get some testable images.

ignore-for-release

Our release process uses GitHub’s release tools, which automatically populate release notes with all PRs that have merged since a baseline version. If your PR will not affect the built containers in any way (for example, it only modifies documentation or unit tests), you should mark it with the ignore-for-release GitHub label to filter it out of the release notes. This can be done at any time before the release process, but it’s simplest to do it when the PR is created.

Internal code changes such as refactoring still change what ends up in the Docker builds, and do not qualify for ignore-for-release.