How to beginner · 4 min read

How to use few-shot prompt template in LangChain

Q: How to use few-shot prompt template in LangChain

Use the FewShotPromptTemplate class from langchain_core.prompts to create few-shot prompts by providing example prompts and a suffix template. This enables structured few-shot learning with any LLM in LangChain.

Quick answer

Use the FewShotPromptTemplate class from langchain_core.prompts to create few-shot prompts by providing example prompts and a suffix template. This enables structured few-shot learning with any LLM in LangChain.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install langchain_openai>=0.2.0

Setup

Install the required LangChain package and set your OpenAI API key in the environment variables.

Run pip install langchain_openai to install LangChain's OpenAI integration.
Set your API key in your shell: export OPENAI_API_KEY='your_api_key_here' (Linux/macOS) or setx OPENAI_API_KEY "your_api_key_here" (Windows).

bash

pip install langchain_openai

Step by step

This example demonstrates creating a few-shot prompt template with two examples and a suffix prompt, then calling OpenAI's gpt-4o model via LangChain to generate a completion.

python

import os
from langchain_openai import ChatOpenAI
from langchain_core.prompts import (
    FewShotPromptTemplate, 
    PromptTemplate, 
    Example
)

# Define example prompts
examples = [
    Example(
        input="Translate 'Hello' to French.",
        output="Bonjour"
    ),
    Example(
        input="Translate 'Goodbye' to French.",
        output="Au revoir"
    )
]

# Define the suffix template where the user input will be inserted
suffix_template = "Translate the following sentence to French: {input}"

# Create a PromptTemplate for the suffix
suffix_prompt = PromptTemplate(
    input_variables=["input"],
    template=suffix_template
)

# Create the FewShotPromptTemplate
few_shot_prompt = FewShotPromptTemplate(
    examples=examples,
    example_prompt=PromptTemplate(
        input_variables=["input"],
        template="Input: {input}\nOutput: {output}"
    ),
    suffix=suffix_prompt,
    input_variables=["input"],
    example_separator="\n\n"
)

# Format the prompt with a new input
formatted_prompt = few_shot_prompt.format(input="I love programming.")
print("Formatted prompt:\n", formatted_prompt)

# Initialize LangChain OpenAI client
client = ChatOpenAI(model="gpt-4o", temperature=0)

# Call the model with the formatted prompt
response = client.invoke([{"role": "user", "content": formatted_prompt}])
print("\nModel response:\n", response.content)

output

Formatted prompt:
Input: Translate 'Hello' to French.
Output: Bonjour

Input: Translate 'Goodbye' to French.
Output: Au revoir

Translate the following sentence to French: I love programming.

Model response:
J'aime programmer.

Common variations

Use different LLMs by changing model in ChatOpenAI, e.g., gpt-4o-mini.
Use async calls with await client.acall([...]) in an async function.
Customize example formatting by modifying example_prompt templates.
Combine few-shot templates with LangChain chains for complex workflows.

Troubleshooting

If you get an authentication error, verify your OPENAI_API_KEY environment variable is set correctly.
If the prompt formatting is incorrect, check that input_variables match your template placeholders.
For rate limits, reduce request frequency or check your OpenAI usage dashboard.

✅

Key Takeaways

Use FewShotPromptTemplate to structure few-shot examples and suffix prompts in LangChain.
Always define input_variables explicitly to match your prompt placeholders.
Format the prompt before sending it to the LLM client for best control over input.
You can easily swap models or use async calls with LangChain's ChatOpenAI client.
Check environment variables and template syntax carefully to avoid common errors.

Verified 2026-04 · gpt-4o

Verify ↗