Just a thought related to LangGraph tho - I am starting to switch my application from langchain to langgraph as I need more control over the chain and conditional edges. One thing I found it is quite difficult when code LangGraph is - each node does have some dependencies from other nodes when you code, meaning like for tool need to grab the tool from latest tool messages so you need to remember and knowing what state and message was passed till this point, it is very difficult to code and even maintain when you have 5+ nodes. I am wondering if there is a better way to design this so each node code logic are independent and all the logic will be coded in the edge or some places.