The current Nexus implementation provides a set of default injection bindings, as defined here, alongside methods for adding bindings for specific types (e.g., input readers, processors, etc.).
However, extending this functionality to support custom bindings would significantly enhance the framework’s flexibility. Specifically, it would allow users to inject any type of object, broadening the scope of use cases the framework can accommodate.
Example use cases include:
Injection of arbitrary MLOps clients: For instance, integrating clients like Weights and Biases directly into the dependency graph, facilitating seamless tracking, and logging of experiments within the existing application infrastructure.
Injection of ML models: This would enable users to inject models that can be subsequently trained, evaluated, and used for inference within the application context, enhancing modularity and reusability.
The current Nexus implementation provides a set of default injection bindings, as defined here, alongside methods for adding bindings for specific types (e.g., input readers, processors, etc.).
However, extending this functionality to support custom bindings would significantly enhance the framework’s flexibility. Specifically, it would allow users to inject any type of object, broadening the scope of use cases the framework can accommodate.
Example use cases include: