Mastering ChatPromptTemplate: Build Memory‑Enabled Conversational Agents with LangChain

When building conversational applications such as chatbots, handling multi‑turn dialogue history requires a prompt structure that can insert a dynamic list of messages. ChatPromptTemplate is designed for this purpose and works together with a ChatModel as a standard tool.

Core components of ChatPromptTemplate

A ChatPromptTemplate consists of one or more message templates, each representing a role in the conversation:

SystemMessagePromptTemplate : creates system messages that set the AI's role, background, and behavior guidelines.

HumanMessagePromptTemplate : creates user messages, representing the human input.

AIMessagePromptTemplate : creates AI messages and can include few‑shot examples to steer the model toward a specific answer format.

MessagesPlaceholder : the key to handling dialogue history

In multi‑turn conversations we cannot know in advance how many turns will occur, so a placeholder is needed to indicate where a dynamic message list should be inserted. MessagesPlaceholder defines a variable that, at runtime, is replaced by a list of messages (e.g., the chat history stored in memory).

from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.messages import HumanMessage, AIMessage

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    MessagesPlaceholder(variable_name="chat_history"),  # placeholder for history
    ("human", "{input}"),  # current user input
])

# At runtime, "chat_history" is replaced with a list of messages
prompt.format_messages(
    chat_history=[
        HumanMessage(content="Hello!"),
        AIMessage(content="Hello! How can I help you?"),
    ],
    input="What is LangChain?"
)

MessagesPlaceholder

is therefore the core component for building chatbots with memory.

Creating a ChatPromptTemplate

The most common way to create a ChatPromptTemplate is via the from_messages class method, which accepts several input formats:

Tuple list (type, content) :

ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    ("human", "{question}"),
])

List of BaseMessagePromptTemplate objects :

from langchain.prompts import SystemMessagePromptTemplate, HumanMessagePromptTemplate
ChatPromptTemplate.from_messages([
    SystemMessagePromptTemplate.from_template("..."),
    HumanMessagePromptTemplate.from_template("..."),
])

List of BaseMessage objects :

from langchain_core.messages import SystemMessage, HumanMessage
ChatPromptTemplate.from_messages([
    SystemMessage(content="..."),
    HumanMessage(content="{question}"),  # only variables are formatted
])

In the examples of this chapter we focus on using MessagesPlaceholder to manage dynamic dialogue history, which is a crucial step for building real chat applications.

Reference

How to: use few‑shot examples in chat models [1] – https://python.langchain.com/docs/how_to/few_shot_chat_examples