* chore: add top p option in llm-tool * chore: wire up the top p with async generate
* feat: add command r model support