Ask more than 50 questions to an LLM + KeyWords

Dear community,

On a text of mine (~ 16000 tokens), I would like to ask to an LLM around 50 different questions. Afterwards, extract the key words from the retrieved answers.

I would like to ask you about general advise for this Data Process Pipeline.

For the 1st stage, the ~50 questions, what should be better, ChatOpenAI class? Even if I do not want any chatbot? … Or better just LLMChain?

Then, for the 2nd stage KeyLLM?

And the ChatOpenAI or LLMChain nested to the KeyLLM with a SequentialChain?

Thank you very much,