content___part

Chapter 2: Content / Part

In Chapter 1: Client, we learned how to configure the Client to connect to Google's Generative AI APIs. Now that we have a connection, we need to understand how to structure the data we send and receive. This chapter introduces the fundamental data structures for communication: Content and Part.

Motivation and Use Case

Generative models, especially multi-modal ones like Gemini, process information in structured turns. A conversation isn't just a flat string; it involves roles (who is speaking?), different types of data (text, images, function calls/results), and potentially multiple pieces of data within a single turn.

The Content and Part objects provide this necessary structure. Content represents a single message or turn in the conversation, while Part represents individual pieces of data within that turn. This allows for rich interactions beyond simple text prompts.

Central Use Case: Imagine you want to ask a Gemini model to describe an image you provide. You can't just pass a raw image; you need to structure the request clearly indicating the text prompt ("What is in this image?") and the image data itself, both belonging to the 'user' role in a single turn.

# Assuming 'client' is configured as shown in Chapter 1
from google.genai import types
import PIL.Image

# Load an image (replace with your image loading logic)
try:
    img = PIL.Image.open('path/to/your/image.jpg')
except FileNotFoundError:
    print("Please replace 'path/to/your/image.jpg' with an actual image file.")
    exit() # Or handle appropriately

# Explicitly construct Content and Part objects
image_part = types.Part.from_image(img) # Helper to create image part
text_part = types.Part.from_text("Describe this image.")

# Create a Content object for the user's turn
user_content = types.Content(
    role='user',
    parts=[text_part, image_part] # Multi-modal input
)

# Send the structured content to the model
response = client.models.generate_content(
    model='gemini-1.5-flash', # Use an appropriate multi-modal model
    contents=[user_content] # Pass a list of Content objects
)

print(response.text)

This example shows how Content (with role='user') groups multiple Part objects (text_part, image_part) to form a single, multi-modal message turn.