VLM(Vision Chat) Skill

This skill guides the implementation of vision chat functionality using the z-ai-web-dev-sdk package, enabling AI models to understand and respond to images combined with text prompts.

Skills Path

Skill Location: {project_path}/skills/VLM

this skill is located at above path in your project.

Reference Scripts: Example test scripts are available in the {Skill Location}/scripts/ directory for quick testing and reference. See {Skill Location}/scripts/vlm.ts for a working example.

Overview

Vision Chat allows you to build applications that can analyze images, extract information from visual content, and answer questions about images through natural language conversation.

IMPORTANT: z-ai-web-dev-sdk MUST be used in backend code only. Never use it in client-side code.

Prerequisites

The z-ai-web-dev-sdk package is already installed. Import it as shown in the examples below.

CLI Usage (For Simple Tasks)

For simple image analysis tasks, you can use the z-ai CLI instead of writing code. This is ideal for quick image descriptions, testing vision capabilities, or simple automation.

Basic Image Analysis

# Describe an image from URL
z-ai vision --prompt "What's in this image?" --image "https://example.com/photo.jpg"

# Using short options
z-ai vision -p "Describe this image" -i "https://example.com/image.png"

Analyze Local Images

# Analyze a local image file
z-ai vision -p "What objects are in this photo?" -i "./photo.jpg"

# Save response to file
z-ai vision -p "Describe the scene" -i "./landscape.png" -o description.json

Multiple Images

# Analyze multiple images at once
z-ai vision \
  -p "Compare these two images" \
  -i "./photo1.jpg" \
  -i "./photo2.jpg" \
  -o comparison.json

# Multiple images with detailed analysis
z-ai vision \
  --prompt "What are the differences between these images?" \
  --image "https://example.com/before.jpg" \
  --image "https://example.com/after.jpg"

With Thinking (Chain of Thought)

# Enable thinking for complex visual reasoning
z-ai vision \
  -p "Count the number of people in this image and describe their activities" \
  -i "./crowd.jpg" \
  --thinking \
  -o analysis.json

Streaming Output

# Stream the vision analysis
z-ai vision -p "Describe this image in detail" -i "./photo.jpg" --stream

CLI Parameters

• --prompt, -p <text>: Required - Question or instruction about the image(s)
• --image, -i <URL or path>: Optional - Image URL or local file path (can be used multiple times)
• --thinking, -t: Optional - Enable chain-of-thought reasoning (default: disabled)
• --output, -o <path>: Optional - Output file path (JSON format)
• --stream: Optional - Stream the response in real-time

Supported Image Formats

• PNG (.png)
• JPEG (.jpg, .jpeg)
• GIF (.gif)
• WebP (.webp)
• BMP (.bmp)

When to Use CLI vs SDK

Use CLI for:

• Quick image analysis
• Testing vision model capabilities
• One-off image descriptions
• Simple automation scripts

Use SDK for:

• Multi-turn conversations with images
• Dynamic image analysis in applications
• Batch processing with custom logic
• Production applications with complex workflows

Recommended Approach

For better performance and reliability, use base64 encoding to pass images to the model instead of image URLs.

Supported Content Types

The Vision Chat API supports three types of media content:

1. image_url - For Image Files

Use this type for static images (PNG, JPEG, GIF, WebP, etc.)

{
    role: 'user',
    content: [
        { type: 'text', text: prompt },
        { type: 'image_url', image_url: { url: imageUrl } }
    ]
}

2. video_url - For Video Files

Use this type for video content (MP4, AVI, MOV, etc.)

{
    role: 'user',
    content: [
        { type: 'text', text: prompt },
        { type: 'video_url', video_url: { url: videoUrl } }
    ]
}

3. file_url - For Document Files

Use this type for document files (PDF, DOCX, TXT, etc.)

{
    role: 'user',
    content: [
        { type: 'text', text: prompt },
        { type: 'file_url', file_url: { url: fileUrl } }
    ]
}

Note: You can combine multiple content types in a single message. For example, you can include both text and multiple images, or text with both an image and a document.

Basic Vision Chat Implementation

Single Image Analysis

import ZAI from 'z-ai-web-dev-sdk';

async function analyzeImage(imageUrl, question) {
  const zai = await ZAI.create();

  const response = await zai.chat.completions.createVision({
    messages: [
      {
        role: 'user',
        content: [
          {
            type: 'text',
            text: question
          },
          {
            type: 'image_url',
            image_url: {
              url: imageUrl
            }
          }
        ]
      }
    ],
    thinking: { type: 'disabled' }
  });

  return response.choices[0]?.message?.content;
}

// Usage
const result = await analyzeImage(
  'https://example.com/product.jpg',
  'Describe this product in detail'
);
console.log('Analysis:', result);

Multiple Images Analysis

import ZAI from 'z-ai-web-dev-sdk';

async function compareImages(imageUrls, question) {
  const zai = await ZAI.create();

  const content = [
    {
      type: 'text',
      text: question
    },
    ...imageUrls.map(url => ({
      type: 'image_url',
      image_url: { url }
    }))
  ];

  const response = await zai.chat.completions.createVision({
    messages: [
      {
        role: 'user',
        content: content
      }
    ],
    thinking: { type: 'disabled' }
  });

  return response.choices[0]?.message?.content