Improve devops -

[jacksonjacobs1]

Description

Re-running the dev container currently requires an image rebuild when code changes. Additionally, pip is slow.

Implementation notes

1. Extract dependencies from pyproject.toml into requirements.txt file. The txt file will be referenced by the toml file.
   https://stackoverflow.com/questions/73600082/how-to-reference-a-requirements-txt-in-the-pyproject-toml-of-a-setuptools-projec
2. Only copy requirements.txt and node dependencies into the container.
3. Remove pip install -e quickannotator from the Dockerfile and instead tell devcontainers to run this after building the image.
4. Add RUN commands to dockerfile to install and use UV instead of pip and virtualenv.

Acceptance criteria

• Code editing no longer invalidates the cached docker image layer
• uv is used for all python package management

[choosehappy]

consider using poetry or uv directly instead of "extracting" stuff from the toml, both appear to support pyproject.tomls directly: https://docs.astral.sh/uv/configuration/files/

[jacksonjacobs1]

Looks like pip does not yet support installing dependencies without installing the main package:
pypa/pip#11440

So if we go the route of using uv to read from the pyproject.toml, we would be tightly coupling quickannotator to uv (note that uv expects a [tool.uv] section in the toml file).

I think we should avoid this tight coupling.

[choosehappy]

wa wa wa... what about something like pip-compile?

https://pip-tools.readthedocs.io/en/latest/cli/pip-compile/

[jacksonjacobs1]

That'd work! does it come with python? Looks like it doesn't on my end.

[choosehappy]

its not a default package, would need to be installed :-\

pip install pip-tools