Hi @everyone, I’m building a multi-prompt conversational voice bot using LangChain + LangGraph. I’m encountering edge cases where the GPT model hallucinates if the number of turns increases, ignores tool calls, or the output parser reaches token limits.