microsoft/JARVIS ? reverse-engineered prompt

I want to build a system like the JARVIS / HuggingGPT project. The main idea is to use a powerful large language model, like one from OpenAI, to act as a controller or a "brain." This brain should be able to take a complex user request, understand it, and then break it down into a series of smaller tasks.

For each task, it should be smart enough to go out and find the best specialized AI model from the Hugging Face community to solve it. After it picks the models, it should execute them, gather the results, and then use the main LLM again to combine everything into a single, coherent answer for the user.

Basically, it's an LLM that can use other AI models as its tools. Let's build it in Python, and I'd like to be able to interact with it through a simple command-line interface to start.