agnes-image-gen

name	agnes-image-gen
description	Agnes AI image generation using agnes-image-2.1-flash (text-to-image) and agnes-image-2.0-flash (image-to-image). Generates images and edits images via the OpenAI-compatible API at https://apihub.agnes-ai.com/v1. Use this skill when the user wants to generate images from text prompts, transform images using references, or perform image editing. Trigger keywords: Agnes AI 生图, agnes-image, agnes image gen, agnes 图片, generate image, 文生图, 图生图, image generation.
agent_created	true

Agnes Image Generation Skill

Generate and edit images using Agnes AI multimodal API. Provides text-to-image and image-to-image capabilities through an OpenAI-compatible interface.

Prerequisites

The user must have an Agnes AI API key. If they do not:

1. Direct them to https://platform.agnes-ai.com/ to register
2. After login, navigate to API Keys and create a new key
3. No credit card required — completely free

API Basics

• Base URL: https://apihub.agnes-ai.com/v1
• Auth: Bearer token via Authorization header
• Compatibility: OpenAI SDK compatible
• Full API docs: Load references/api_docs.md when detailed parameter info is needed

Models

Model	Use
agnes-image-2.1-flash	Text-to-image
agnes-image-2.0-flash	Image-to-image, editing, multi-image composition

Image Generation

Using the Bundled Script (Recommended)

The scripts/generate_image.py script handles both text-to-image and image-to-image:

# Text-to-image (using AGNES_API_KEY env var)
python scripts/generate_image.py \
  --prompt "一只可爱的柴犬在樱花树下睡觉" \
  --size 1024x1024 \
  -o output.png

# Image-to-image (style transfer, editing)
python scripts/generate_image.py \
  --prompt "改成水彩画风格" \
  --image https://example.com/photo.png \
  --size 1024x768 \
  -o output.png

API Key: 脚本会自动读取 AGNES_API_KEY 环境变量,无需手动传入 --api-key。也可以显式传入 --api-key YOUR_KEY 覆盖环境变量。

Direct API Call

When the script cannot be used, make a direct API call:

from openai import OpenAI

client = OpenAI(
    api_key="API_KEY",
    base_url="https://apihub.agnes-ai.com/v1"
)

# Text-to-image — DO NOT pass extra_body
response = client.images.generate(
    model="agnes-image-2.1-flash",
    prompt="prompt text",
    size="1024x1024"
)
image_url = response.data[0].url

# Image-to-image — MUST pass extra_body
response = client.images.generate(
    model="agnes-image-2.0-flash",
    prompt="transformation description",
    size="1024x768",
    extra_body={
        "tags": ["img2img"],
        "image": ["https://example.com/source.png"],
        "response_format": "url"
    }
)

Key Rules for Image Generation

• Text-to-image: use agnes-image-2.1-flash, do NOT include extra_body
• Image-to-image: use agnes-image-2.0-flash, MUST include extra_body with tags=["img2img"]
• Chinese prompts are fully supported
• Supported sizes: 1024x1024, 1024x768, 768x1024, 512x512

Workflow

When the user asks to generate an image or edit an image:

1. Scripts automatically read AGNES_API_KEY env var — no need to pass --api-key
2. Determine if it is text-to-image, image-to-image, or image editing
3. For image generation: run scripts/generate_image.py with appropriate flags (no --api-key needed if env var is set)
4. For image-to-image: run scripts/generate_image.py --image URL --prompt description
5. After generation, deliver the output file to the user

If direct API calls are preferred (e.g., within a Python environment), use the OpenAI SDK as shown above. Load references/api_docs.md for comprehensive parameter details, response formats, and troubleshooting known issues.