For example, I told the Agent my name in the first round of conversation. After starting a new conversation, when I ask the Agent my name, it can’t answer. Now I want to pass the previous chat history before each round of conversation; is this possible?
Add a session variable node, select append, and append past chat history in each turn.
Do I need to build a chatflow and then add variable nodes to it?
No, just add one more node, “variable assignment”, to concatenate each conversation record and place it before the agent. For example, if the first round’s input is “I am Xiao Ming”, and the second input is “Please act as my friend…”, then the second round’s model or agent node (context selects variable) will receive the prompt “I am Xiao Ming, please act as my friend…”.
现在是把记忆功能开关打开就行了，agent的话，不清楚你们那时候有没有agent啊。llm那个上下文还只能用一个变量，难绷