[RFE] Universal LLM Parser that can handle slight deviations

[shawn-hurley]

LLM's responses can vary, and sometimes they don't follow the expected format. We generally have similar patterns for the responses and how we want to parse them. We should make this generic so that new agents can use them.