workflow.py

from langchain.schema import BaseMemory
from langchain.prompts import PromptTemplate
from langchain.memory import ConversationBufferMemory
from langchain_core.runnables.history import RunnableWithMessageHistory
from config import get_chat_model

def get_session_history(memory: BaseMemory):
    """Retrieve the session history from memory."""
    return memory.chat_memory.messages if hasattr(memory, "chat_memory") else []

def get_workflow():
    """Set up the chatbot workflow with memory and prompt template."""

    # Define the prompt template for both general conversation and weather retrieval
    prompt_template = PromptTemplate(
        input_variables=["input", "previous_conversation"],
        template="""
You are a helpful assistant. You should answer the user's question or have a normal conversation. If the user asks about the weather, 
please respond with the current weather information based on their input location. Otherwise, answer to the best of your ability.

If the user's input is about the weather, you should respond with details about the weather.
For example:
- "What is the weather in Paris?"
- "How's the weather in New York?"

Example conversation flow:
User: What's the weather like today in London?
Assistant: Let me check the weather for you. The current weather in London is [weather details].

If the input is not weather-related, just respond with a conversational response.

The user said: {input}
Previous conversation: {previous_conversation}
""")
    
    # Create memory for conversation
    memory = ConversationBufferMemory(memory_key="input", return_messages=True)

    # Fetch the chat model
    chat_model = get_chat_model()

    # Use RunnableWithMessageHistory for session memory
    conversation_chain = RunnableWithMessageHistory(
        runnable=chat_model,
        get_session_history=lambda: get_session_history(memory),
        memory=memory,
        verbose=True,
        prompt=prompt_template
    )

    return conversation_chain