Why? The inference server isn't a harness, it's tokens in, tokens out. That's different from a harness.