Chat History & MemoryLesson 2

InMemoryChatMessageHistory

The last lesson kept history in a plain list and appended HumanMessage / AIMessage by hand. InMemoryChatMessageHistory does the same job with simpler calls: add_user_message and add_ai_message. Pass chat_history.messages to format_messages — same steps as ChatPromptTemplate, less code.

Plain list vs helper

The prompt template and model call do not change. You only replace how you save each turn.

Previous lesson — plain list

history = []

history.append(HumanMessage(...))
history.append(AIMessage(...))

InMemoryChatMessageHistory

chat_history = InMemoryChatMessageHistory()

chat_history.add_user_message(...)
chat_history.add_ai_message(...)
Same turns stored either way. InMemoryChatMessageHistory gives you helper methods instead of building message objects yourself. Later lessons swap in Redis or SQL with the same API.

Set it up

Import from langchain_core.chat_history. One object is enough for this demo.

from langchain_core.chat_history import InMemoryChatMessageHistory

chat_history = InMemoryChatMessageHistory()

Methods you will use most:

.messages— pass to format_messages as history
.add_user_message(text)— save the question
.add_ai_message(text)— save the reply
.clear()— wipe stored turns
.messages holds the same HumanMessage and AIMessage objects the template expects.

Each turn

Read chat_history.messages, call the model, then save the question and reply. Same four steps as the previous lesson.

messages = prompt.format_messages(
    history=chat_history.messages,
    input=question,
)
reply = model.invoke(messages)

chat_history.add_user_message(question)
chat_history.add_ai_message(reply.content)

The demo script

chat_message_history_demo.py runs the same two HTML <a> questions — now with InMemoryChatMessageHistory.

◇chat_message_history_demo.py
"""chat_message_history_demo.py"""
chat_history = InMemoryChatMessageHistory()
history=chat_history.messages
add_user_message · add_ai_message

Download the code

chat_message_history_demo.py

Two HTML questions — uses InMemoryChatMessageHistory

Download .py
Uses langchain_core — already included with the course packages. Venv and .env from Project Setup.

Run it

Activate the venv from Project Setup, then:

python chat_message_history_demo.py
PowerShell — (.venv) active
(.venv) PS C:\projects\langchain-course> python chat_message_history_demo.py
Human: What does the HTML <a> tag do?
AI: The <a> tag creates a hyperlink to another page or resource.
Human: What attribute opens the link in a new tab?
AI: Use target="_blank" on the <a> tag.
Stored 4 messages in chat_history.
Four stored messages — two questions and two replies. The second answer works because the first exchange is in chat_history.messages.

Quick reference

  • InMemoryChatMessageHistory — lives in memory for this script run. Use Redis or SQL when you need data after a restart.
  • .messages — use instead of the plain history list in format_messages.
  • .clear() — empty history when starting over.
  • Docs: Message history · InMemoryChatMessageHistory API.

What's Next

Next: RunnableWithMessageHistory — load and save history per session_id without the manual loop.