Vision Analyze

Analyze images using the built-in vision capabilities of multimodal AI models.

Quick Start

Analyze an Image

Describe what's in an image:

# The agent will automatically use vision when you provide an image path
image("/path/to/image.jpg", prompt="Describe what's in this image")

Extract Text (OCR)

Extract text from images:

image("/path/to/document.png", prompt="Extract all text from this image")

Analyze Multiple Images

Compare or analyze multiple images:

images(["/path/to/image1.jpg", "/path/to/image2.jpg"], 
       prompt="Compare these two images and describe the differences")

Usage Patterns

Visual Q&A

Ask specific questions about image content:

image("menu.jpg", prompt="What are the prices of the main courses?")
image("chart.png", prompt="What trend does this graph show?")
image("screenshot.png", prompt="What error message is displayed?")

Content Moderation

Check image content:

image("upload.jpg", prompt="Is this image appropriate for a professional setting?")

Data Extraction

Extract structured data from visual content:

image("receipt.jpg", prompt="Extract the date, total amount, and items purchased")
image("business_card.png", prompt="Extract name, phone, email, and company")
image("form.jpg", prompt="Extract all filled fields as key-value pairs")

Visual Comparison

Compare images:

images(["before.jpg", "after.jpg"], 
       prompt="What changes were made between these two images?")

Tips

• Be specific: The more specific your prompt, the better the results
• Multiple images: You can analyze up to 20 images at once
• Supported formats: JPG, PNG, GIF, WebP
• Size limits: Large images are automatically resized

When to Use

• Reading text from screenshots, documents, or photos
• Describing visual content for accessibility
• Analyzing charts, graphs, or diagrams
• Comparing visual changes
• Extracting data from forms or receipts
• Understanding UI elements or error messages