LangChain Memory & State: Building Context-Aware Conversations

🤖 Why Context is Critical

Out of the box, LLMs forget everything after each message. That's fine for isolated prompts, but frustrating in real conversations.

LangChain's memory modules fix this. They let you track past exchanges, maintain continuity, and build smarter chat systems.

🛠️ Memory Types in LangChain

LangChain supports different memory strategies:

• ConversationBufferMemory – Saves the full conversation as-is
• SummaryMemory – Condenses chat history into a brief summary
• EntityMemory – Tracks facts about specific people, places, or objects

For this chapter, we'll use ConversationBufferMemory.

⚙️ Step-by-Step: Adding Memory

from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationChain
from langchain.llms import GooglePalm

llm = GooglePalm()

memory = ConversationBufferMemory()

chain = ConversationChain(
    llm=llm,
    memory=memory,
    verbose=True
)

chain.run("Hi, I'm Mark.")
chain.run("What's my name?")

This will return:

"You just told me your name is Mark."

Boom — your app now has memory.

🧩 When to Use Memory

• Multi-turn chat
• Smart assistants
• Customer support bots
• Workflow builders that ask follow-up questions

LangChain memory brings a human touch to LLM interactions.

🔜 Next Up: Adding Tools to Your LLM

In Chapter 4, we'll add tool use — so your LLM can calculate, search, or hit APIs as part of its reasoning.

👉 Continue to Chapter 4: Using LangChain Tools — Add Web Search, Math, and Custom Actions