Zai Image

Name: Zai Image
Rating: 3.5 (1 reviews)
Author: yakinrei

By yakinrei 👁 54 views ▲ 0 votes

Z.AI CogView-4 and GLM-Image generation plugin for OpenClaw

GitHub

Configuration Example

{
     "plugins": {
       "entries": {
         "zai-image": { "enabled": true }
       },
       "allow": ["zai-image"],
       "load": {
         "paths": [
           "/path/to/.openclaw/plugins/zai-image"
         ]
       }
     }
   }

README

# Z.AI Image Generation Plugin for OpenClaw

An [OpenClaw](https://openclaw.ai) code plugin that adds Z.AI's image generation models (**CogView-4** and **GLM-Image**) as a first-class image generation provider.

## Why This Plugin Exists

OpenClaw ships with a built-in Z.AI provider, but it only covers **text/chat completions** (GLM-5.2, GLM-5-Turbo, etc.). The image generation models — [CogView-4](https://docs.z.ai/guides/image/cogview-4) and [GLM-Image](https://docs.z.ai/guides/image/glm-image) — are not included, because they use a separate API endpoint (`/paas/v4/images/generations`) with a different request/response format.

This plugin bridges that gap by registering a dedicated `zai-image` image generation provider that talks directly to the Z.AI Images API.

## Features

- ✅ Supports **GLM-Image**, **CogView-4** (`cogview-4-250304`), **CogView-4** (legacy), and **CogView-3 Flash**
- ✅ Automatic image download (Z.AI returns URLs, plugin fetches and converts to buffers)
- ✅ Reads existing Z.AI API key from `models.providers.zai` config — no extra setup if you already use Z.AI for chat
- ✅ Quality mapping (`hd` / `standard`)
- ✅ Size hints per model
- ❌ No image editing (the Z.AI API does not support edits)

## Installation

### Option A: Local plugin directory

1. Clone or copy this folder to your plugins directory:

   ```bash
   git clone https://github.com/yakinrei/openclaw-zai-image.git ~/.openclaw/plugins/zai-image
   ```

2. Add to your `openclaw.json`:

   ```json
   {
     "plugins": {
       "entries": {
         "zai-image": { "enabled": true }
       },
       "allow": ["zai-image"],
       "load": {
         "paths": [
           "/path/to/.openclaw/plugins/zai-image"
         ]
       }
     }
   }
   ```

3. (Optional) Set as your default image generation model:

   ```json
   {
     "imageGenerationModel": {
       "primary": "zai-image/cogview-4-250304"
     }
   }
   ```

4. Restart your gateway:

   ```bash
   openclaw gateway restart
   ```

### Option B: ClawHub

```bash
clawhub package install yakinrei/zai-image
```

## Configuration

The plugin reads your existing Z.AI API key from `models.providers.zai.apiKey` in `openclaw.json`. If that's not set, it falls back to the `ZAI_API_KEY` environment variable.

**Important:** The plugin uses the **general API endpoint** (`https://api.z.ai/api/paas/v4`), not the Coding Plan endpoint (`/api/coding/paas/v4`). This is because image generation is billed separately from the Coding Plan.

## ⚠️ Limitations & Important Notes

### Z.AI Coding Plans Do NOT Include Image Generation

The **GLM Coding Plan** (Lite / Pro / Max) is a subscription **for coding only** — it covers text models like GLM-5.2 and GLM-5-Turbo via the `/api/coding/paas/v4` endpoint.

**Image generation is pay-as-you-go** and billed per image:

| Model       | Price per Image |
|-------------|-----------------|
| GLM-Image   | $0.015          |
| CogView-4   | $0.01           |

To use this plugin, you need to **add balance (credits) to your Z.AI account** at [z.ai/manage-apikey/billing](https://z.ai/manage-apikey/billing). The Coding Plan subscription quota cannot be used for image generation.

### Other Limitations

- **Single image per request** — the Z.AI API currently returns one image per call
- **No image editing** — Z.AI's image endpoint does not support reference image editing
- **URL expiry** — generated image URLs expire after 30 days; the plugin downloads them immediately so this is not an issue for local storage
- **Content filtering** — Z.AI enforces content safety filters; requests that violate their policies will be rejected
- **Quality differences** — `glm-image` defaults to `hd` quality (~20s generation), while `cogview-4` defaults to `standard` (~5-10s)

## Supported Models

| Model ID              | Label              | Default Size  | Notes                        |
|-----------------------|--------------------|---------------|------------------------------|
| `cogview-4-250304`    | CogView-4          | 1024x1024     | Latest CogView-4 release     |
| `glm-image`           | GLM-Image          | 1280x1280     | Higher quality, slower       |
| `cogview-4`           | CogView-4 (legacy) | 1024x1024     | Legacy alias                 |
| `cogview-3-flash`     | CogView-3 Flash    | 1024x1024     | Fastest, lower quality       |

## API Reference

This plugin calls the [Z.AI Generate Image API](https://docs.z.ai/api-reference/image/generate-image):

```
POST https://api.z.ai/api/paas/v4/images/generations
Authorization: Bearer <your-api-key>
Content-Type: application/json

{
  "model": "cogview-4-250304",
  "prompt": "A cute kitten on a windowsill",
  "size": "1024x1024"
}
```

Response:

```json
{
  "created": 1760335349,
  "data": [{ "url": "https://..." }]
}
```

## License

MIT

## Author

[@yakinrei](https://github.com/yakinrei)

tools