How to Create Prompt Template in Langchain: Simple Guide

In Langchain, create a prompt template using the PromptTemplate class by defining a template string with placeholders and specifying input variables. This lets you easily generate dynamic prompts by filling in values for those variables.

Syntax

The PromptTemplate class requires two main parts: a template string with placeholders wrapped in curly braces, and a list of input_variables that match those placeholders. When you call format() on the template, you provide values for these variables to create the final prompt string.

python

from langchain.prompts import PromptTemplate

prompt = PromptTemplate(
    input_variables=["name", "age"],
    template="Hello, my name is {name} and I am {age} years old."
)

final_prompt = prompt.format(name="Alice", age=30)
print(final_prompt)

Output

Hello, my name is Alice and I am 30 years old.

Example

This example shows how to create a prompt template that asks for a user's favorite color and hobby. It then formats the prompt by filling in those values.

python

from langchain.prompts import PromptTemplate

# Define the prompt template with placeholders
prompt = PromptTemplate(
    input_variables=["color", "hobby"],
    template="What do you think about the color {color} when you are {hobby}?"
)

# Fill in the placeholders with actual values
filled_prompt = prompt.format(color="blue", hobby="painting")
print(filled_prompt)

Output

What do you think about the color blue when you are painting?

Common Pitfalls

Common mistakes include:

• Not matching input_variables names exactly with placeholders in the template string.
• Forgetting to provide all required variables when calling format().
• Using incorrect placeholder syntax (must be {variable}).

Here is an example of a wrong and right way:

python

# Wrong: variable name mismatch
from langchain.prompts import PromptTemplate

try:
    prompt_wrong = PromptTemplate(
        input_variables=["username"],
        template="Hello, {name}!"
    )
    prompt_wrong.format(username="Bob")
except KeyError as e:
    print(f"Error: {e}")

# Right: matching variable names
prompt_right = PromptTemplate(
    input_variables=["name"],
    template="Hello, {name}!"
)
print(prompt_right.format(name="Bob"))

Output

Error: 'name'
Hello, Bob!

Quick Reference

Remember these key points when creating prompt templates in Langchain:

• Use PromptTemplate with template and input_variables.
• Placeholders in the template must be wrapped in curly braces {}.
• Call format() with all required variables to get the final prompt string.

Key Takeaways

Use PromptTemplate with matching input_variables and placeholders to create dynamic prompts.

Always wrap placeholders in curly braces {} inside the template string.

Provide all required variables when calling format() to avoid errors.

Check variable names carefully to prevent KeyError exceptions.

PromptTemplate helps keep prompt creation clean and reusable.