Notebook metadata editor widget

[MaxTuecke]

Adds #475

Overview

Adds display_metadata_widget(), which displays an ipywidget form that lets users easily edit a notebook's metadata.

Discussion

In place of the old top cell containing the commented out metadata json, the new top cell of a notebook will import and run display_metadata_widget().

If the user makes any changes to their notebooks requirements, the widget will show a 'Install new requirements' button that installs the new requirements to the container and updates your local requirements file if one was provided.

When a user updates their notebook's metadata with the widget, it will immediately json save the new NotebookMetadata object to /garden/notebook_metadata.json

In order to then save the metadata to the notebook, jupyter is now being launched with a custom config that adds a post save hook. When the user saves their notebook, the hook goes and looks for the json NotebookMetadata file and saves the garden metadata with the rest of the notebook's metadata. Jupyter can get upset about file versions if we just used nbformat to add the metadata, so the post save hook is the best way to edit the notebook's file while the notebook is running.

Other changes:

• build_notebook_session_image now copies in custom_jupyter_config.py to the container so jupyter can be launched with the custom config. The last line of this file c.FileContentsManager.post_save_hook = post_save_hook is appended to the file during runtime, since jupyter's magic variable c causes an error when trying to import the script.
• start_container_with_notebook has a new argument, custom_config, which is default True. This adds --config=/garden/custom_jupyter_config.py to the entrypoint command for the container. The only time this would be turned off is when using notebook debug. Otherwise, the post save hooks should always be turned on.
• start_container_with_notebook now launches the container with a new env vars NOTEBOOK_PATH and REQUIREMENTS_PATH which is the notebook's and requirements (if one was provided) path inside the container. It's a pain to get the either while it is running, so this just makes life easier.
• start_container_with_notebook is binding the users requirements file (if one was provided) to the one in the container.
• The 'sister' python script that is made during build_notebook_session_image will exclude the tagged widget cell from the script.
• Changed NotebookMetadata and RequirementsData from namedtuple to pydantic BaseModel.
• notebook start will now always try to add the widget cell and empty garden_metadata dict to the notebook if either is missing. The widget cell is tagged with garden_display_metadata_cell
• Updated the notebook templates with the new format.

Testing

Updated the old test_metadata tests to test for the new format.

---

📚 Documentation preview 📚: https://garden-ai--482.org.readthedocs.build/en/482/

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 36.84211% with 108 lines in your changes missing coverage. Please review.

Project coverage is 75.50%. Comparing base (9ccdbd0) to head (4c2cb03).

Files	Patch %	Lines
garden_ai/notebook_metadata.py	27.73%	86 Missing :warning:
garden_ai/scripts/custom_jupyter_config.py	31.25%	11 Missing :warning:
garden_ai/app/notebook.py	0.00%	9 Missing :warning:
garden_ai/containers.py	88.23%	2 Missing :warning:

:exclamation: Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #482 +/- ## ========================================== - Coverage 77.42% 75.50% -1.92% ========================================== Files 54 55 +1 Lines 3592 3695 +103 ========================================== + Hits 2781 2790 +9 - Misses 811 905 +94 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[MaxTuecke]

As we talked about in person, I added in a 'Install new requirements' button that pops up when any changes are made to the requirements w/ the widget. It will install new reqs, update the users local reqs file (if one was provided) and then restart the jupyter kernel.

I also added info popup messages to help the user understand how to use the widget.

As Owen / Hayden suggested I also:

• Changed NotebookMetadata and RequirementsData to pydantic BaseModel
• Changed the metadata saving from pickle to json
• Removed the conda functionality for the widget since we are planning on removing that functionality anyways.