Metadata

name: agnes-image
description: Generate or edit images using Agnes Image 2.1 Flash via OpenAI-compatible images API. Supports text-to-image and image-to-image (img2img) generation. Use this skill when the user asks to: generate an image, create AI art, draw/paint with AI, edit an image with AI, do image-to-image translation, use Agnes for image generation. Triggers: 生成图片, AI绘画, 文生图, 图生图, Agnes图片, agnes-image, generate image, create picture.
agent_created: true

Agnes Image — Agnes Image 2.1 Flash 图像模型

Overview

Agnes Image 2.1 Flash is Sapiens AI's free image generation model with an OpenAI-compatible images API. Supports text-to-image generation and image-to-image editing.

Model ID: agnes-image-2.1-flash
Base URL: https://apihub.agnes-ai.com/v1
Pricing: ~$0.003/image (free tier available)
Auth: Authorization: Bearer YOUR_API_KEY
Generate endpoint: POST /v1/images/generations

Quick Start

API key from AGNES_API_KEY environment variable.

Python (OpenAI SDK)

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AGNES_API_KEY"],
    base_url="https://apihub.agnes-ai.com/v1",
)

# Text-to-Image
resp = client.images.generate(
    model="agnes-image-2.1-flash",
    prompt="A cute cat playing soccer on a green field, sunny day, cinematic style",
    size="1024x1024",
)
print(resp.data[0].url)

cURL (Text-to-Image)

curl https://apihub.agnes-ai.com/v1/images/generations \
  -H "Authorization: Bearer $AGNES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agnes-image-2.1-flash",
    "prompt": "A futuristic city at sunrise, cinematic realism",
    "size": "1024x768",
    "extra_body": {"response_format": "url"}
  }'

cURL (Image-to-Image)

curl https://apihub.agnes-ai.com/v1/images/generations \
  -H "Authorization: Bearer $AGNES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agnes-image-2.1-flash",
    "prompt": "Turn the scene into a rainy cyberpunk night with neon lights",
    "size": "1024x768",
    "extra_body": {
      "image": ["https://example.com/input.png"],
      "response_format": "url"
    }
  }'

Endpoint

Item	Value
Method	`POST`
URL	`https://apihub.agnes-ai.com/v1/images/generations`
Content-Type	`application/json`
Auth Header	`Authorization: Bearer <key>`

Request Parameters

Parameter	Type	Required	Description
`model`	string	Yes	`agnes-image-2.1-flash`
`prompt`	string	Yes	Image description / edit instruction
`size`	string	Yes	Output size, e.g. `1024x768`, `1024x1024`
`image`	string[]	For img2img	Input image URLs or Data URIs
`return_base64`	boolean	No	Return base64 instead of URL
`extra_body`	object	No	Extra parameters (see below)
`extra_body.response_format`	string	No	`url` (default) or `b64_json`
`extra_body.image`	string[]	For img2img	Alternative way to pass input images

Response Format

{
  "created": 1780000000,
  "data": [
    {
      "url": "https://storage.googleapis.com/agnes-aigc/xxx.png",
      "b64_json": null,
      "revised_prompt": null
    }
  ]
}

Access the image at data[0].url. For base64 output, use data[0].b64_json.

Capabilities

1. Text-to-Image (文生图)

Generate an image from a text description. Use detailed prompts with:

[主体] + [场景/环境] + [风格] + [光线] + [构图] + [质量要求]

Example: A futuristic city marketplace, flying vehicles, holographic signs, dense crowds, neon lighting, cinematic realism, ultra-detailed

2. Image-to-Image (图生图)

Edit an existing image while preserving its structure. Provide the input image URL via extra_body.image:

edit = client.images.generate(
    model="agnes-image-2.1-flash",
    prompt="Keep the cat and soccer ball, change scene to snowy winter field",
    size="1024x1024",
    extra_body={"image": ["https://example.com/original.png"]},
)

3. Base64 Output

For programmatic processing without downloading from URL:

resp = client.images.generate(
    model="agnes-image-2.1-flash",
    prompt="A simple red circle on white background",
    size="512x512",
    return_base64=True,
)
b64_data = resp.data[0].b64_json

Important Notes

response_format must be placed inside extra_body, NOT at the top level
Image generation may take 10-60 seconds; set client timeout to 60-360s
Input images for img2img must be publicly accessible HTTPS URLs or Data URIs
Do NOT add tags: ["img2img"] — just provide the image array
Both agnes-image-2.1-flash (text-to-image) and agnes-image-2.0-flash (image-to-image focused) are available; 2.1 handles both tasks

Bundled Scripts

`scripts/generate.py`

Generate images from the command line:

# Text-to-image
python scripts/generate.py --prompt "A dragon flying over mountains" --size 1024x768

# Image-to-image
python scripts/generate.py --prompt "Make it nighttime" --input-image https://example.com/photo.png

# Save to file
python scripts/generate.py --prompt "Sunset beach" --output sunset.png

Reads AGNES_API_KEY from environment.

Reference

See references/api_reference.md for complete parameter details, error handling, prompt engineering tips, and more examples.

Agnes Image 2.0 Flash

安装技能