Best way to run an API with LLM?

[xrd]

I'm interested in using JavaScript with the llm tool. I could use spawn and just capture the output but I'm wondering if restarting the llm each time is a bad idea and if it would be better to talk to llm from my JavaScript with an API. I'm fine with hard coding the parameters and model and don't need much more than the ability to ask questions and get back answers. Should I use an additional python library or does llm bundle this kind of functionality at all?

[xrd]

This looks good. I'll try some of this.

#4