Not storing prompts but they absolutely can store reasoning traces, summarized or sanitized reasoning traces and the whole LLM output as well.
Not storing prompts but they absolutely can store reasoning traces, summarized or sanitized reasoning traces and the whole LLM output as well.