You are tasked with maintaining a machine learning workflow for your organization, which involves a Vertex AI ML pipeline. This pipeline includes preprocessing and training steps, each running on separate custom Docker images. Your organization uses GitHub for version control and GitHub Actions as the CI/CD tool for running unit and integration tests. There is a need to automate the model retraining process so that it can be triggered manually or automatically when a new version of the code is merged into the main branch. The goal is to configure the CI/CD workflow in a way that minimizes the steps required and offers maximum flexibility. How should you set up this CI/CD workflow?