During release wave 1, a AI cloud developer owns a agent orchestration gateway. The main requirement is to host a Python inference API from a private container image while injecting endpoint settings as App Service environment variables. What should be implemented?