pi-minimax-image

MiniMax Image Understanding extension for pi coding agent — provides AI-powered image understanding as a fallback for models without native vision support.

⚠️ Use only when the model does not natively support image understanding. Many models (Claude, Gemini, GPT-4V) can process images directly via the read tool. This extension's understand_image tool routes images through the MiniMax API — use it only when the model you're running cannot natively see images.

Why This Package?

Built for image understanding fallback. While pi-web-access handles web search and URL fetching, this package specializes in AI-powered image analysis for models without built-in vision.

Use this package when you need to:

• 📸 Analyze screenshots, diagrams, and UI mockups
• 🔤 Extract text from images (OCR)
• 🖥️ Transcribe code from screenshots
• 🎨 Get AI descriptions of visual content
• 📊 Analyze charts, graphs, and data visualizations

No conflicts. This extension only provides understand_image - it won't interfere with other packages like pi-web-access.

Features

• 🖼️ Image Understanding - Analyze images with AI for descriptions, OCR, code extraction, and visual analysis
• ⚡ Easy Configuration - Configure via environment variables or pi settings files
• 🔄 Hot Reload - Changes apply without restarting pi
• 🎨 Rich UI - Custom rendering with progress indicators and status updates
• ✅ Focused - Only adds image understanding, no bloat

Prerequisites

• pi coding agent installed
• MiniMax Coding Plan subscription
• Node.js >= 18.0.0

Installation

pi install npm:@zgltyq/pi-minimax-image

Or via git:

pi install git:github.com/zgltyq/minimax-understand-image

Configuration

Get Your API Key

1. Visit MiniMax Coding Plan to subscribe
2. Get your API key from API Key page

Environment Variable

export MINIMAX_API_KEY="your-api-key-here"

Auth File

Save to ~/.pi/agent/auth.json:

{
  "minimax": {
    "type": "api_key",
    "key": "your-api-key-here"
  }
}

Interactive Setup

/minimax-configure --key your-api-key-here

Usage

Image Understanding

understand_image({
  prompt: "What error is shown in this screenshot?",
  image_url: "https://example.com/error.png"
})

Image Sources

• HTTP/HTTPS URLs: "https://example.com/image.jpg"
• Local paths: "/home/user/image.png" or "./screenshot.jpg"

Example Use Cases

// Describe image content
understand_image({
  prompt: "Describe what's in this image in detail",
  image_url: "https://example.com/photo.jpg"
})

// Extract text (OCR)
understand_image({
  prompt: "Extract all text from this image",
  image_url: "./screenshots/document.png"
})

// Analyze UI/UX
understand_image({
  prompt: "Analyze this UI design and suggest improvements",
  image_url: "https://example.com/mockup.png"
})

// Code from screenshot
understand_image({
  prompt: "What code is shown in this screenshot? Transcribe it exactly.",
  image_url: "./error-screenshot.jpg"
})

// Analyze data visualizations
understand_image({
  prompt: "What trends and insights can you extract from this chart?",
  image_url: "./analytics.png"
})

Tool Reference

understand_image

| Parameter | Type | Required | Description | | --------- | ------ | -------- | ------------------------------------------------ | | prompt | string | ✓ | Question or analysis request (1-1000 characters) | | image_url | string | ✓ | Image URL or local file path |

Commands

| Command | Description | | -------------------- | -------------------------- | | /minimax-configure | Configure API key | | /minimax-status | Show configuration status |

Related Packages

• pi-web-access - Web search, URL fetching, GitHub repos, YouTube videos, PDFs

Learn More

• MiniMax Coding Plan
• pi coding agent

License

MIT