kedacore/keda ? reverse-engineered prompt

Build me a Kubernetes based event driven autoscaling tool that can scale workloads up and down, including all the way to zero, based on external events like queue length or message traffic.

I want it to feel like a real operator for Kubernetes, so it should work with the normal autoscaling flow, expose metrics in a way Kubernetes can use, and let users define scaling rules through a custom resource. It should be something I can deploy with YAML or Helm, and it should run cleanly in a cluster without needing a bunch of extra services.

Please include a solid default setup, basic docs for getting started, and a few example scalers or sample workloads so someone can try it quickly. If you need to check current Kubernetes docs or best practices while building it, go ahead and look them up online.