fnnx-ai/scikit-llm ? reverse-engineered prompt

Build me a Python library that makes it easy to use large language models like ChatGPT inside a normal scikit learn workflow for text tasks.

I want something where I can set my OpenAI key once, load a text dataset, then create a classifier that feels like a regular scikit learn model, so I can call fit and predict on text data without having to write a lot of custom LLM code myself. A simple zero shot text classification example would be great, with a small demo dataset and clear README instructions so someone can try it quickly.

Make it clean, easy to install, and structured like a real package. If you need current API details or best practices, look them up online first. Also include tests and enough documentation so it’s obvious how to use it for text analysis tasks.