from operator import itemgetter

from langchain_core.runnables import RunnablePassthrough

chain = (
    RunnablePassthrough.assign(messages=itemgetter("messages") | trimmer)
    | prompt
    | model
)

response = chain.invoke(
    {
        "messages": messages + [HumanMessage(content="what's my name?")],
        "language": "English",
    }
)

The above code is from the LangChain documentation for creating a chatbot with memory. They use trim_messages to reduce the messages to fit the context window which is understandable.

But, then they wrap this chain with RunnableWithMessageHistory which doesn’t make sense to me using:

with_message_history = RunnableWithMessageHistory(
    chain,
    get_session_history,
    input_messages_key="messages",
)

config = {"configurable": {"session_id": "abc20"}}

response = with_message_history.invoke(
    {
        "messages": messages + [HumanMessage(content="whats my name?")],
        "language": "English",
    },
    config=config,
)

response.content

Considering the fact that RunnableWithMessageHistory is supposed to automatically log and fetch messages previously sent, why do they have a messages key in the dictionary being passed to invoke method? I am very very confused as to what’s going on over here. Please explain to me how the bottom code works. If it is wrong, how can I integrate trim_messages to automatically fetch messages from history.

I am just confused. I don’t even know what to try.