I’m working with LangChain to generate responses based on user inputs. My initial prompt requires some variables, and I use these variables to generate a response. The problem arises when I try to ask a follow-up question based on the initial response; the model throws errors because it expects the initial variables again.

How can I manage the context so that follow-up questions can be asked without needing to re-input the initial variables?

This is my code.

class Chat:
    def __init__(self):
        self.llm = ChatOpenAI(
            api_key=os.getenv('OPENAI_API_KEY'),
            model='gpt-4o',
            temperature=0.3,
        )
        self.parser = JsonOutputParser(pydantic_object=TargetingType)
        self.prompt_template = ChatPromptTemplate.from_messages([
            ("system", SYSTEM_PROMPT_ONE),
            ("human", USER_PROMPT),
        ])
    def get_chain(self):
        return self.prompt_template | self.llm | self.parser

MAIN FILE

chain = self.chat.get_chain()
response = chain.invoke({
  // input variables
})

// do some work on the response
// After everything is done I want to feed again the data to the Model which is different from the initial request