Welcome to the first lesson of our "Building an Image Generation Service With FastAPI" course! In this course, you'll learn how to build a complete web application that transforms text descriptions into stunning images using Google's Gemini model through the Gemini API.

Before we dive into the FastAPI framework or API integration, we need to establish a solid foundation for our image generation system. At the heart of any AI image generation service is the prompt — the text instructions that guide the AI in creating the image you want.

Prompt engineering is the art and science of crafting effective instructions for AI models. When working with native image generation models like Gemini, the quality and structure of your prompts directly impact the quality of the images you receive. A well-crafted prompt provides clear direction, specific details, and appropriate context to help the AI understand exactly what you're looking for.

In our application, we'll be creating event banners for a fictional company called "Eventify Co." Rather than crafting a new prompt each time a user requests an image, we'll create a template system that:

1. Maintains consistent structure and quality across all prompts
2. Allows users to customize only the specific event details
3. Handles the formatting and presentation of the prompt automatically

This approach ensures that our application produces high-quality, consistent results while still allowing for customization. Let's begin by understanding what makes an effective prompt template.

A well-structured prompt template for image generation typically contains several key components that work together to guide the AI. Let's examine the structure of our template:

# ROLE
Lead graphic designer

# THEME 
Use this exact event text: "{user_input}"

# TASK
Create a visually stunning, high-quality 16:9 event banner for the event described above.
Prominently include only the exact event text from the THEME section. Avoid adding any other text.
Ensure the text is readable, well-integrated with the design, and visually balanced.

# DESIGN REQUIREMENTS
- Color Palette: Muted gold, deep navy blue, charcoal black, ivory white.
- Style: Sophisticated pastel shades and metallic accents conveying exclusivity and elegance.
- Typography: 
  - Event names: Elegant serif fonts.
  - Taglines or details: Sleek sans-serif fonts.
- Composition: Visual clarity and balance with generous spacing and harmonious layout.

# OUTPUT REQUIREMENTS
The banner must be suitable for:
- Social media
- Websites
- Print

Ensure high resolution and polished visual quality. Maintain brand consistency and deliver an impactful design ready for promotional use.

Let's break down each section:

ROLE: This establishes the persona that the AI should adopt. By positioning the AI as a lead graphic designer, we're setting expectations for high-quality, professional output.

THEME: This is where we'll insert the user's input — the specific event details they want to feature in the banner. Notice the {user_input} placeholder, which we'll programmatically replace with actual content.

TASK: This section clearly defines what we want the AI to create — an event banner with specific characteristics. It provides direction on how text should be integrated into the design.

DESIGN REQUIREMENTS: Here we provide specific design guidelines, including color palette, style, typography, and composition. These details help ensure consistency across all generated images and align with the brand identity of our fictional company.

OUTPUT REQUIREMENTS: This final section specifies the practical requirements for the image, ensuring it will be suitable for various use cases.

By structuring our prompt this way, we're providing comprehensive guidance to the AI while still allowing for customization through the user input.

Now that we understand the structure of our prompt template, let's create the actual file that will store it. In our application, we'll organize files according to their function, with templates stored in a dedicated directory.

First, let's set up our project directory structure:

app/
├── data/
│   └── image_prompt_template.txt
├── models/
│   └── prompt_manager.py
└── main.py

The data directory will store our template and potentially other data files. The models directory will contain our Python classes, and main.py will be our entry point for testing.

Now, let's create the image_prompt_template.txt file in the data directory with the content we discussed in the previous section. You can use any text editor to create this file and save it with the exact structure we reviewed earlier.

Make sure the file is saved with UTF-8 encoding to handle any special characters properly. The placeholder {user_input} is crucial — this is what allows our code to dynamically insert the user's specific event details into the template.

If you're working in a team environment, consider adding comments at the top of the file explaining its purpose and how it should be modified.

With our template file in place, we now need a way to load it and format it with user input. For this, we'll create a PromptManager class that handles these operations. This class will be responsible for:

1. Loading the template from the file
2. Inserting user input into the template
3. Handling any errors that might occur during these operations

Let's create the prompt_manager.py file in the models directory:

class PromptManager:
    """Manages the loading and formatting of the prompt template for consistency."""
    
    @classmethod
    def load_base_prompt(cls, file_path: str = 'data/image_prompt_template.txt') -> str:
        """Load the base prompt from a file."""
        try:
            with open(file_path, 'r') as f:
                return f.read()
        except Exception as e:
            print(f"Error loading prompt template: {e}")
            return (
                "# ROLE\nYou are a professional graphic designer.\n\n"
                "# THEME\nUse this exact event text: '{user_input}'\n\n"
                "# TASK\nPlease generate a beautiful 16:9 banner.\n"
            )

    @classmethod
    def format_prompt(cls, user_input: str) -> str:
        """Format the prompt by inserting the user input into the template."""
        base_prompt = cls.load_base_prompt()
        return base_prompt.format(user_input=user_input)

The PromptManager class has two class methods, which means they can be called directly on the class without creating an instance. The @classmethod decorator is used to define these methods. This is appropriate here since we're not storing any state specific to an instance; the methods operate on the class level.

Notice the error handling in load_base_prompt. This is important because it ensures our application won't crash if the template file is missing or corrupted. Instead, it will fall back to a simplified template that can still produce reasonable results.

Now that we have our template file and PromptManager class, let's create a simple script to test that everything works correctly. We'll create a main.py file in the root of our app directory:

from models.prompt_manager import PromptManager

user_input = "Luxury Tech Conference 2025: Innovating the Future - April 10th, New York City"
formatted_prompt = PromptManager.format_prompt(user_input)

print("Generated Prompt:")
print(formatted_prompt)

This script:

1. Imports our PromptManager class
2. Defines a sample user input for a fictional tech conference
3. Calls the format_prompt method to insert this input into our template
4. Prints the resulting formatted prompt

When you run this script, the output will look like this:

Generated Prompt:
# ROLE
Lead graphic designer

# THEME 
Use this exact event text: "Luxury Tech Conference 2025: Innovating the Future - April 10th, New York City"

# TASK
Create a visually stunning, high-quality 16:9 event banner for the event described above.
Prominently include only the exact event text from the THEME section. Avoid adding any other text.
Ensure the text is readable, well-integrated with the design, and visually balanced.

As you can see, our user input has been successfully inserted into the THEME section of the template. The rest of the template remains unchanged, providing consistent guidance to the Gemini AI model.

In this lesson, we built the foundation for our image generation service by designing a reusable and resilient prompt template system. You now have the tools to load a template, insert dynamic content, and gracefully handle errors.

Next, we'll continue building our service by creating an ImageManager class to manage the storage and retrieval of generated images, preparing us to handle the native image format returned by Gemini 3.1 Flash.