README.md

• Run ray start --head in another directory (simulating remote environment in K8s)
• Run python app.py, it will sync the current directory, including my_module.py to the remote cluster
• curl localhost:8000/f returns this from my_module.py::my_func

app.py

import ray
from ray import serve

ray.init("ray://localhost:10001", runtime_env={"working_dir": "."}, namespace="serve")

@serve.deployment
def f(_):
    from my_module import my_func
    return my_func()

serve.start(detached=True)
f.deploy()

my_module.py

def my_func():
    return "this from my_module.py::my_func"