Image Generation

Image Generation Skill

Overview

Z.ai's glm-image API generates images from text prompts. OpenAI-compatible endpoint at https://api.z.ai/api/paas/v4/images/generations.

Requirements:

• API key in Z_AI_API_KEY environment variable
• Use xh (preferred) or curl with Content-Type: application/json header

Quick Start

# Generate image (xh - recommended)
xh --ignore-stdin POST https://api.z.ai/api/paas/v4/images/generations \
  Authorization:"Bearer $Z_AI_API_KEY" \
  Content-Type:application/json \
  model=glm-image \
  prompt="A cute kitten on a sunny windowsill" \
  size=1280x1280

# Generate image (curl)
curl -s -X POST https://api.z.ai/api/paas/v4/images/generations \
  -H "Authorization: Bearer $Z_AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-image",
    "prompt": "A cute kitten on a sunny windowsill",
    "size": "1280x1280"
  }'

Response Format

{
  "created": 1774317025,
  "data": [{"url": "https://...watermark.png?UCloudPublicKey=..."}],
  "id": "...",
  "request_id": "..."
}

Image URL is temporary (expires after some time). Download immediately if persistence needed.

Post-Generation Workflow

After receiving the image URL, always download the image to a local path (default /tmp).

# Download to /tmp, with a unique representitive name
curl -sL "$IMAGE_URL" -o /tmp/xxx.png

Then read the image using the Read tool, to confirm image correctness.

Parameters

Extract Image URL

# xh + jq
xh --ignore-stdin POST https://api.z.ai/api/paas/v4/images/generations \
  Authorization:"Bearer $Z_AI_API_KEY" \
  Content-Type:application/json \
  model=glm-image \
  prompt="..." \
  size=1280x1280 | jq -r '.data[0].url'

# Download directly
xh --ignore-stdin POST ... | jq -r '.data[0].url' | xargs curl -o image.png

Error Handling

Common errors:

• 401 Unauthorized: Invalid or missing API key
• 400 Bad Request: Invalid parameters (check size format, prompt length)

Always check HTTP status code and response body for error details.

Money Budget

Z.ai's glm-image is an expensive model with following budgets:

• Cost: 40 RMB
• Limit: 500 requests per-month

Each invocation to image models are precious and expensive. Keep in mind that each invocation consumes user's money for about 0.08 RMB.

Constraints:

• Do NOT re-run with exact same prompt - re-use the previous generated image URL.
• Do NOT send repetitive invocations of the image generation endpoint for debugging - use GET endpoints (e.g. listing models) to test connection instead.