When drawing diagrams, you can draw them either from the entire system or just it's part perspective. Of course it is normal to use both approaches in a single project. It all depends what is your goal and what are you trying to model.
When looking at the FE, you can of course draw at least following diagrams:
- Use case diagram to indicate what functions are offered by the FE. It will often be similar to UC diagram of the entire system, however it will definitely not contain the APIs exposed by the system for other systems to integrate as those are BE specific. Besides, you may have different front-ends (e.g. fat vs thin client), offering different functionalities
- Activity diagram to show flow (e.g. screens' flow, e.g. per UC)
- Class diagram to show data structures used by FE (they often differ to some level to the BE part, however usually are generated by the FE-BE integration layer)
- Sequence diagram or Communication diagram to indicate points of integration with BE, especially in complex cases
Of course there can be much more, depending on your specific situation (e.g. timing diagram can be important in your specific case)
Having said that, take into account that you don't need to draw any diagrams. The idea of modelling is to draw those diagrams, that are needed, i.e. help grasp some ideas that may be otherwise be difficult to understand.