Add a simple LLM workflow that uses the output structure generated by #1 in order to answer questions.
