🧾 Lesson: Using Prompt Templates

Module: Building a Basic AI Agent


🎯 Lesson Objective

By the end of this lesson, you’ll be able to:

  • Understand what a prompt template is in LangChain

  • Create reusable templates to structure AI input

  • Dynamically fill in prompts with user queries

  • Use templates to guide the model’s behavior more precisely


🧠 What Is a Prompt Template?

A prompt template is a structured input string that:

  • Controls how you speak to the LLM

  • Makes AI outputs more reliable, focused, and customizable

  • Allows for variables (e.g. {question} or {context}) that are filled dynamically at runtime

Think of it as a “smart sentence with blanks” that you fill in depending on what you’re asking.


📦 Step-by-Step: Using Prompt Templates in LangChain


✅ 1. Import PromptTemplate

from langchain.prompts import PromptTemplate

✅ 2. Define a Template

Here’s a simple one for question answering:

template = """
You are a helpful and intelligent assistant.

Answer the question below clearly and concisely.

Question: {question}

Answer:
"""

prompt = PromptTemplate(
    input_variables=["question"],
    template=template
)

You’ve now created a reusable prompt that accepts a single variable: question.


✅ 3. Format the Prompt

Use prompt.format() to fill in the variable:

formatted_prompt = prompt.format(question="What is quantum computing?")
print(formatted_prompt)

🖨️ Output:

You are a helpful and intelligent assistant.

Answer the question below clearly and concisely.

Question: What is quantum computing?

Answer:

✅ 4. Use with Your Model

Pass the formatted prompt to your LLM:

from langchain.llms import Ollama

llm = Ollama(model="mistral", temperature=0.5)

response = llm(formatted_prompt)
print("🤖", response)

🧠 Why Use Prompt Templates?

Feature Benefit
🧩 Structure Ensure input follows a logical format
🧠 Role assignment Guide the model (e.g. “You’re a tutor”)
🔁 Reusability Easy to reuse with different questions
🎯 Control Reduces randomness and vague answers

💡 Examples of Useful Templates

📘 Example 1: Tutor Bot

template = """
You are an English teacher. Explain the following concept clearly for a beginner student.

Concept: {concept}

Explanation:
"""

🧮 Example 2: Math Solver

template = """
You are a smart math tutor. Show the steps clearly and explain the answer.

Problem: {problem}

Solution:
"""

💬 Example 3: Role-based Persona

template = """
You are an AI assistant that answers like a pirate.

Question: {question}

Answer like a pirate:
"""

🧠 Advanced: Chain Templates with Chains

PromptTemplates are usually used inside LLMChain:

from langchain.chains import LLMChain

qa_chain = LLMChain(
    llm=llm,
    prompt=prompt
)

response = qa_chain.run("What is the difference between RAM and ROM?")
print(response)

✅ Summary

Step What You Did
🧾 Created a prompt template With variables like {question}
🧠 Used .format() to insert user input Dynamically filled templates
🧠 Sent the prompt to the LLM To get structured, context-aware responses
🔁 Combined with LLMChain For reusability and modular AI behavior

 

47