Few-shot prompt templates in LangChain - Mini Project: Build & Apply

Create a Few-Shot Prompt Template with Langchain

📖 Scenario: You are building a chatbot that answers questions by learning from a few example conversations. You want to create a prompt template that includes these examples to guide the AI's responses.

🎯 Goal: Build a few-shot prompt template using Langchain that includes example questions and answers, and a placeholder for the user's question.

📋 What You'll Learn

Create a list of example conversations as dictionaries with 'question' and 'answer' keys

Define a prefix string that introduces the examples

Use Langchain's FewShotPromptTemplate to combine the prefix, examples, and a suffix with a placeholder

Set up the final prompt template that can be used to generate prompts with new questions

💡 Why This Matters

🌍 Real World

Few-shot prompt templates help AI models understand how to answer questions by showing a few examples first, similar to teaching by example in real life.

💼 Career

Knowing how to build few-shot prompts is useful for AI developers and prompt engineers who create conversational agents and smart assistants.

1

Set up example conversations data

Create a list called examples with two dictionaries. The first dictionary should have 'question': 'What is the capital of France?' and 'answer': 'Paris'. The second dictionary should have 'question': 'Who wrote Hamlet?' and 'answer': 'William Shakespeare'.

LangChain

# Create the list of example conversations
# Your code here

Need a hint?

Use a list with two dictionaries exactly as described.

2

Define the prefix string for the prompt

Create a string variable called prefix with the exact value: 'Answer the following questions based on the examples:'.

LangChain

examples = [
    {'question': 'What is the capital of France?', 'answer': 'Paris'},
    {'question': 'Who wrote Hamlet?', 'answer': 'William Shakespeare'}
]
# Define the prefix string
# Your code here

Need a hint?

Assign the exact string to the variable prefix.

3

Create the FewShotPromptTemplate with Langchain

Import FewShotPromptTemplate and PromptTemplate from langchain.prompts. Then create a PromptTemplate called example_prompt with input variables question and answer, and template string Q: {question}\nA: {answer}. Next, create a FewShotPromptTemplate called few_shot_prompt using examples, example_prompt, prefix, and a suffix string Q: {input}\nA: with input variable input.

LangChain

examples = [
    {'question': 'What is the capital of France?', 'answer': 'Paris'},
    {'question': 'Who wrote Hamlet?', 'answer': 'William Shakespeare'}
]
prefix = 'Answer the following questions based on the examples:'

# Import FewShotPromptTemplate and PromptTemplate
# Create example_prompt with question and answer variables
# Create few_shot_prompt using examples, example_prompt, prefix, and suffix
# Your code here

Need a hint?

Use the exact class names and variable names as specified.

4

Use the few-shot prompt template with a new question

Create a variable called new_question with the value 'What is the boiling point of water?'. Then create a variable called final_prompt by calling few_shot_prompt.format(input=new_question).

LangChain

from langchain.prompts import FewShotPromptTemplate, PromptTemplate

examples = [
    {'question': 'What is the capital of France?', 'answer': 'Paris'},
    {'question': 'Who wrote Hamlet?', 'answer': 'William Shakespeare'}
]
prefix = 'Answer the following questions based on the examples:'

example_prompt = PromptTemplate(
    input_variables=['question', 'answer'],
    template='Q: {question}\nA: {answer}'
)

few_shot_prompt = FewShotPromptTemplate(
    examples=examples,
    example_prompt=example_prompt,
    prefix=prefix,
    suffix='Q: {input}\nA:',
    input_variables=['input']
)

# Create new_question variable
# Format final_prompt using few_shot_prompt.format with input=new_question
# Your code here

Need a hint?

Use the exact variable names and call the format method with the correct argument.