feat: improve serve commands and expose DYNAMO_HOME env var

[jon-chuang]

Overview:

/workspace only exists on dockerized deployment.

Details:

Expose DYNAMO_HOME which is configured to /workspace within provided containers, and user-configurable for non-docker deployments

One may prefer a non-dockerized dev setup since even changing one line of python code results in 2 minutes of build time

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

👋 Hi jon-chuang! Thank you for contributing to ai-dynamo/dynamo.

Just a reminder: The NVIDIA Test Github Validation CI runs an essential subset of the testing framework to quickly catch errors.Your PR reviewers may elect to test the changes comprehensively before approving your changes.

🚀

[ryanolson]

/workspace is usually the name of our working directory inside a container.

We should update the instructions to use a DYNAMO_HOME env, then set that env in the container to point at where ever our WORKDIR in the image is set.

Thanks again for helping!

[nnshah1]

One may prefer a non-dockerized dev setup since even changing one line of python code results in 2 minutes of build time

One other option for local dev is also to mount into the container and pip install the packages as editable:

./container/run.sh --mount-workspace -it
cd /workspace/lib/bindings/python
uv pip install -e .
...

[jon-chuang]

One other option for local dev is also to mount into the container and pip install the packages as editable:

Cool. I think that is preferable because the VLLM patches and NIXL build are pain points that can be avoided with that workflow.

Would be cool if there is first class support for that

[jon-chuang]

I saw this PR added instructions for it. #463
Glad to see it!