Truss handle should reflect changes made to truss after it was created

[pankajroark]

Typical flow is like this:

handle = mk_truss(model, target_directory='some_dir')
# Make changes to truss, like modify `model.py`
# Recreate handle to load changes
handle = from_directory('some_dir')

That last step should is easy to forget and should not be needed. It will be great if the handle automatically reflects changes tot he underlying truss.

[bolasim]

In my experience, modifying model.py and re-runing handle.predict works with the now code withouth needing to call truss.load("...").

Is there something else that we want to make sure if reflected (maybe configs? requirements?)

[pankajroark]

When we create the truss handle we also create the truss_spec, which reads in the config. This config in truss_handle would go out of sync when the config.yaml is modified on disk. We'd need to orchestrate things, so that there's some kind of revalidation/reload when any truss_handle methods are called.

[bolasim]

Ah interesting. Do we have to run a kind of filewatcher, like FastAPI does for fast reload?

[bolasim]

Do we still need this as we move to a more cli focused flow, over a notebook focused flow?