Memory & Chat History

from standins import RuleModel
from langchain.messages import SystemMessage, HumanMessage, AIMessage, trim_messages
from langchain_core.chat_history import InMemoryChatMessageHistory
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.runnables.history import RunnableWithMessageHistory

class EchoMemory(RuleModel):
    """Stand-in that can only 'remember' what is inside the messages it receives - exactly like a real LLM."""
    def _pick(self, messages):
        seen = " ".join(str(m.content) for m in messages[:-1])
        q = str(messages[-1].content).lower()
        if "name" in q or "summarise" in q:
            facts = [f for f in ["Ajeet", "Pune", "3-day trip", "Rs 15,000", "mountains"] if f.lower() in seen.lower()]
            return "You told me: " + ", ".join(facts) if facts else "Sorry, I don't know that - you haven't told me yet."
        return "Noted!"
llm = EchoMemory()

print("no memory :", llm.invoke("What is my name and where do I live?").content)

chat_history = [SystemMessage("You are a friendly assistant. Keep answers short.")]
def chat(user_input):
    chat_history.append(HumanMessage(user_input))
    response = llm.invoke(chat_history)
    chat_history.append(AIMessage(response.content))
    return response.content
chat("Hi, my name is Ajeet and I live in Pune.")
print("manual    :", chat("What is my name and where do I live?"), f"({len(chat_history)} messages sent)")

prompt = ChatPromptTemplate.from_messages([("system", "You are a helpful travel assistant."), MessagesPlaceholder("history"), ("human", "{input}")])
store = {}
def get_session_history(session_id):
    if session_id not in store: store[session_id] = InMemoryChatMessageHistory()
    return store[session_id]
chatbot = RunnableWithMessageHistory(prompt | llm, get_session_history, input_messages_key="input", history_messages_key="history")
cfg = {"configurable": {"session_id": "ajeet"}}
chatbot.invoke({"input": "I want a 3-day trip from Pune with Rs 15,000 budget."}, config=cfg)
chatbot.invoke({"input": "I like mountains."}, config=cfg)
print("ajeet     :", chatbot.invoke({"input": "Summarise my requirements."}, config=cfg).content)
print("priya     :", chatbot.invoke({"input": "Summarise my requirements."}, config={"configurable": {"session_id": "priya"}}).content)
print("store     :", {k: len(v.messages) for k, v in store.items()})

trimmed = trim_messages(store["ajeet"].messages, max_tokens=12, strategy="last", token_counter=lambda ms: sum(len(str(m.content).split()) for m in ms), start_on="human")
print("trimmed   :", len(store["ajeet"].messages), "->", len(trimmed), [m.content[:30] for m in trimmed])