import os from langchain_core.messages import HumanMessage, AIMessage from dotenv import load_dotenv from langchain_google_genai import ChatGoogleGenerativeAI load_dotenv() # Memory is handled by passing a list of messages back and forth chat_history = [ HumanMessage(content="My name is Dnyan."), AIMessage(content="Nice to meet you, Dnyan! How can I assist with your emails?") ] # The agent now 'remembers' the context follow_up = "What is my name and can you draft a thank you note?" llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash") response = llm.invoke(chat_history + [HumanMessage(content=follow_up)]) print(response.content) # The agent will correctly identify your name from the history