Ideally, we could even use a OpenRouter endpoint for Game Text Prompt, and a local model (LLM Studio) for the rest of them. This would be a powerful solution. That being said, even though I would only want to have one endpoint for Game Text, and a different one for the other 3 prompts; probably other users could enjoy having different endpoints for all the 4 prompts and I hope it wouldn't increase the complexity on your end much, adding 3 more endpoints versus adding only 1 more.
If you are concerned about simplicity of use, you could have a checkbox: "Use only one endpoint" as a default. Then if you uncheck it, a dropdown allows the user to define the rest of the endpoints.
Honestly, I don't know how this should interact with the API keys. In my case I would use the same for all of them, as LLM Studio requires no API key...