Skip to main content
Learn how to elevate your prompt engineering by turning messages into reusable LangChain prompt templates. This guide shows you how to create, populate, and invoke chat prompts with the OpenAI API using ChatPromptTemplate, SystemMessagePromptTemplate, and HumanMessagePromptTemplate.

​
Prerequisites and Setup

Ensure you have the required packages installed: 
pip install langchain-openai
Import the necessary modules and set your OpenAI API key: 
from langchain_openai import ChatOpenAI
from langchain.prompts import (
    HumanMessagePromptTemplate,
    SystemMessagePromptTemplate,
    ChatPromptTemplate
)
import os

# Make sure your API key is set, for example:
# export OPENAI_API_KEY="your_api_key"
Never commit your API key to version control. Use environment variables or a secrets manager.

​
Choosing the Right Prompt Classes

LangChain provides several classes to structure your chat prompts. Use the table below as a quick reference:
ClassPurposeExample
ChatPromptTemplateCombine system & human messagesChatPromptTemplate.from_messages([...])
SystemMessagePromptTemplateDefine system-level instructionsSystemMessagePromptTemplate(prompt=...)
HumanMessagePromptTemplateTemplate for user-level messagesHumanMessagePromptTemplate(prompt=...)
For more details, see the LangChain Prompt Templates documentation.

​
1. Defining Reusable Message Templates

Start by creating message strings with placeholders: 
# Template strings with placeholders
sys_msg   = "You are a {subject} teacher"
human_msg = "Tell me about {concept}"
These templates aren’t tied to concrete values yet—they’re blueprints for your prompts.

​
2. Building a ChatPromptTemplate

Use ChatPromptTemplate.from_messages to stitch together your system and human templates: 
prompt_template = ChatPromptTemplate.from_messages([
    ("system", sys_msg),
    ("human",  human_msg)
])
At this stage, prompt_template holds the structure but lacks the actual subject and concept values.

​
3. Populating and Formatting the Prompt

Fill in the placeholders at runtime: 
filled_messages = prompt_template.format_messages(
    subject="Chemistry",
    concept="Periodic Table"
)
Now filled_messages contains two messages:
  • System: “You are a Chemistry teacher”
  • Human: “Tell me about Periodic Table”

​
4. Invoking the Chat Model

Send the formatted messages to the OpenAI chat model: 
model    = ChatOpenAI()
response = model.invoke(filled_messages)
print(response.content)
Typical output:
The Periodic Table is a tabular arrangement of the chemical elements, organized by atomic number, electron configuration, and recurring chemical properties. Rows are called periods, columns are groups, and elements in the same group share similar chemical behaviors.

​
5. Extending with Few-Shot Prompting

You can embed few-shot examples or custom tones by explicitly composing templates: 
from langchain.prompts import PromptTemplate

prompt_template = ChatPromptTemplate(
    input_variables=["subject", "concept"],
    messages=[
        SystemMessagePromptTemplate(
            prompt=PromptTemplate(
                input_variables=["subject"],
                template="You are a {subject} teacher"
            )
        ),
        HumanMessagePromptTemplate(
            prompt=PromptTemplate(
                input_variables=["concept"],
                template="Tell me about {concept}"
            )
        )
    ]
)

filled = prompt_template.format_messages(
    subject="Physics",
    concept="Quantum Mechanics"
)

model    = ChatOpenAI()
response = model.invoke(filled)
print(response.content)
This pattern makes it easy to add example exchanges or change the model’s persona.
Few-shot prompting can increase token usage—and cost—so monitor your token counts and model billing.

​
References

Watch Video

Practice Lab