Description
Add highlight to kedro-viz section of dataset.metadata to highlight it in the UI as a "key dataset" - e.g. with yellow background fill and black text.
Context
I'm regularly walking the team / audience through the kedro viz UI and almost always, on this huge DAG representation there are 2-3 datasets which are really the key artifacts I explain the flow in respect to.
Usually those are:
- Master table
- Trained model
Using modular pipelines, and hence leveraging the collapsing supernodes, is incredibly helpful is simplifying the DAG view and visually tracking where those key artifacts are. However, I though that since the metadata injection logic already exists in the kedro viz (e.g. adding layer to datasets), I thought that if the group decides that this feature is useful, it might be relatively easy to implement.
Possible Implementation
So the most elegant way I see is to do this exactly like e.g. layer can be added to a dataset in its catalog definition.
For example, let's say the trained_model is the one I want to highlight. See where I'm adding highlight: true:
# catalog.yml
trained_model:
type: ...
filepath ...
metadata:
kedro-viz:
layer: models
highlight: true
Possible Alternatives
Probably the above is the most natural place for this functionality to live, so not sure.
Checklist
Description
Add
highlighttokedro-vizsection ofdataset.metadatato highlight it in the UI as a "key dataset" - e.g. with yellow background fill and black text.Context
I'm regularly walking the team / audience through the
kedro vizUI and almost always, on this huge DAG representation there are 2-3 datasets which are really the key artifacts I explain the flow in respect to.Usually those are:
Using modular pipelines, and hence leveraging the collapsing supernodes, is incredibly helpful is simplifying the DAG view and visually tracking where those key artifacts are. However, I though that since the metadata injection logic already exists in the
kedro viz(e.g. addinglayerto datasets), I thought that if the group decides that this feature is useful, it might be relatively easy to implement.Possible Implementation
So the most elegant way I see is to do this exactly like e.g.
layercan be added to a dataset in itscatalogdefinition.For example, let's say the
trained_modelis the one I want to highlight. See where I'm addinghighlight: true:Possible Alternatives
Probably the above is the most natural place for this functionality to live, so not sure.
Checklist