I had to memorize scikit import paths for tfidf tokenizers. Things are different now Codex by chatgpt by codex app by chat by OpenAI