Prompt Templates

When you build an app, you don't rewrite the prompt for every user. You create a Template.

The Concept

A template is a string with variables.

• Static: "Write a poem about a cat."
• Dynamic: "Write a poem about a {{animal}}."

In Python (f-strings)

The simplest template is a Python f-string.

def make_poem_prompt(animal, emotion):
    return f"""
    You are a poet.
    Write a short poem about a {animal}.
    The poem must evoke the emotion of {emotion}.
    """
    
prompt = make_poem_prompt("Cat", "Anger")
# Send 'prompt' to Gemini

Advanced Templating (LangChain)

If you use libraries like LangChain, templates get more powerful.

• Partials: You can pre-fill some variables (e.g., set style="Shakespeare") and leave others open for the user.
• Validation: Ensure inputs aren't too long or empty.

AI Studio Variables

In AI Studio's "Structured Prompt", the input columns act as variables. When you click "Get Code," it generates a loop that handles the injection of data into these slots.

Summary

• Never hardcode user data.
• Use variables ({}) to inject dynamic content.
• Sanitize inputs (check for length) before injecting to save money and safety.

Module 4 Complete! You are now a Prompt Engineer. In Module 5, we go beyond prompting: Fine-Tuning and Custom Models.