my_memory_logic.py

# my_memory_logic.py

import os

# Import the "run_with_chain_context" function from pipeline.py
# This function must accept a dict with { "input": ..., "chat_history": ... }
# and return a dict with { "answer": ... }.
from pipeline import run_with_chain_context

# For session-based chat history
from langchain_core.chat_history import BaseChatMessageHistory
from langchain_community.chat_message_histories import ChatMessageHistory
from langchain_core.runnables.history import RunnableWithMessageHistory

###############################################################################
# 1) In-Memory Store: session_id -> ChatMessageHistory
###############################################################################
store = {}  # e.g., { "abc123": ChatMessageHistory(...) }

def get_session_history(session_id: str) -> BaseChatMessageHistory:
    """
    Retrieve (or create) a ChatMessageHistory for the given session_id.
    This ensures each session_id has its own conversation transcripts.
    """
    if session_id not in store:
        store[session_id] = ChatMessageHistory()
    return store[session_id]

###############################################################################
# 2) Build a RunnableWithMessageHistory that wraps "run_with_chain_context"
###############################################################################
# "run_with_chain_context" must be a function returning a dict,
# e.g. { "answer": "... final string ..." }
# input_messages_key -> "input"
# history_messages_key -> "chat_history"
# output_messages_key -> "answer"

conversational_rag_chain = RunnableWithMessageHistory(
    run_with_chain_context,    # from pipeline.py
    get_session_history,
    input_messages_key="input",
    history_messages_key="chat_history",
    output_messages_key="answer"
)

###############################################################################
# 3) A convenience function that calls our chain with session-based memory
###############################################################################
def run_with_session_memory(user_query: str, session_id: str) -> str:
    """
    Calls the 'conversational_rag_chain' with a given session_id and user_query.
    This returns the final 'answer' from run_with_chain_context.
    """
    response = conversational_rag_chain.invoke(
        {"input": user_query},
        config={
            "configurable": {
                "session_id": session_id
            }
        }
    )
    return response["answer"]