feature: custom callback functionality in vllm-router

[pwuersch]

Describe the feature

This feature would allow custom python code to be triggered on events such as a finished request or other points in the code. It would be something similar to LiteLLM's Custom Callback functionality.

Why do you need this feature?

We want to add some custom code to scan the returned code snippets for plagiarism, amongst other stuff. Later down the line we would also need to add some custom authentication logic to the different endpoints provided by the vllm-router.

Additional context

We would be happy to contribute such a feature if this aligns with your vision for the project overall.

[YuhanLiu11]

Hey @pwuersch thanks for submitting the issue! Can you elaborate the custom authentication logic you would like to add to the endpoints?