tobiasuhmann / rui-fe Goto Github PK

When starting RUI via docker-compose, API calls don't work because the backend container is only reachable from within the Docker network.

As a quickfix, publish the backend container's port 5000.

Currently, clicking an item in the taxonomy selects it and toggles its expansion status, i.e. expands it if collapsed and vice versa.

Instead, clicking the item should only select it, while clicking the +/- icon on its left should expand/collapse it. To facilitate clicking the icon, its size should be increased.

On the predictions page, the current node (specified in the URL) is highlighted in the taxonomy:

In addition, the taxonomy should also highlight nodes that are part of other predictions. The highlighting should be different. Note: This might require showing further root nodes and updating the highlighting when the predictions page is switched.

The user should be aware where he is within the application. On the upload and taxonomy pages, the respective headers are bold. On the predictions page, this is not the case:

Add breadcrumbs that show the current location within the app and allow navigating "upwards" intuitively.

Currently, the loading indicator layer covers the whole page:

Instead, the loading indicator should only cover the main content.

Also, check that handling page switches while loading something is handled correctly when navigating via the then active header. (This is already possible by changing the URL while loading).

Links shall be styled as buttons:

Currently, the Docker image is built in the pipeline. In addition, it should be tested whether it can be deployed.

Currently, the backend returns the paged predictions in the order they were inserted:

Instead, the most promising predictions should be shown at the top. Therefore, a score shall be calculated for each prediction as follows:

1. The top normalized synonym and child scores should be averaged. If no synonym or child prediction is available, only the normalized child or synonym score is taken, respectively.
2. If multiple entries have the same total normalized score, the non-normalized score (calculated the same way as the normalized score) is taken.

The total normalized and non-normalized scores shall be displayed in the to-be-created header of the prediction cards. The scores shall be calculated in the backend to keep pagination possible.

The rui-fe pipeline is broken. Unit and e2e do not pass on main. Also, the Docker deployment tests do not pass.

As usual, clicking the top-left (RUI) logo should lead to the home page, i.e. the taxonomy page in case of RUI.

On the predictions page, when selecting tokens for a prediction, the proposed mention updates respectively. However, after editing the proposed mention manually, the mention is not updated anymore. This was intended but should be changed. The proposed mention should always update to match the selected tokens when the token selection changes.

Example: Selecting "mollis" after editing the proposed mention

Currently, there is no direct feedback when triggering an action like adding an entity. This easily leads to the user performing the action multiple times.

Therefore, a loading indicator should be implemented that blocks further UI interaction and tells the user what he is currently waiting for. The loading indicator should only appear if the action takes longer than 100ms.

The user should immediately recognize elements he can interact with, such as items in the taxonomy as well as clickable tokens on the predictions page. Currently, the cursor changes to a pointer when hovering over those elements:

In addition, the respective element should be highlighted by changing its color. Also, fix the behavior for non-clickable taxonomy items - the cursor should not change for those.