[Feature] Better prompt

[smstevemoretz1]

Describe the issue/feature here!

This is a follow up to: https://github.com/voideditor/void/issues/820

I figured for all the IDEs such as cursor that support custom models you can just configure a server or even a local server with nginx and just listen to the incoming requests.

Obviously this is for educational purposes and Void team can use this method to enhance void a lot.

For a simple node.js server is enough for getting the initial request, but you can proxy it back to open.ai and make it really work and get the next prompts too to analyize everything, this is probably the easiest way to do this but it can be done with manual LLMs too if you have a good hardware.

https://www.tensorzero.com/blog/reverse-engineering-cursors-llm-client/

Anyway I can help you with the node.js proxy part if you need me let me know.