Image generation

The Otoroshi LLM extension provides a unified, OpenAI-compatible API for generating and editing images across multiple providers. You can configure image generation providers, expose them through Otoroshi routes, and let consumers generate images using a standard API.

Supported providers

Provider	Generation	Editing	Default model
OpenAI	Yes	Yes	gpt-image-1
Azure OpenAI	Yes	No	gpt-image-1
Cloud Temple	Yes	Yes	—
Gemini	Yes	Yes	imagen-3.0-generate-002
Grok (X-AI)	Yes	No	grok-2-image
Luma	Yes	No	photon-1
Hive	Yes	No	black-forest-labs/flux-schnell

Features

• Image generation from text prompts (all providers)
• Image editing from uploaded images + text prompt (OpenAI, Gemini, Cloud Temple)
• Model routing — route to different providers using provider/model syntax in the model field
• Decode mode — return raw image bytes instead of JSON when generating a single image
• Model constraints — restrict which models consumers can use via include/exclude regex patterns, enforceable per API key or per user
• Budget enforcement — image generation costs are tracked and budgets are enforced
• Cost tracking — per-request cost tracking integrated with cost tracking

API endpoints

The extension exposes two OpenAI-compatible endpoints through Otoroshi route plugins:

Endpoint	Method	Description
/v1/images/generations	POST	Generate images from a text prompt
/v1/images/edits	POST	Edit an existing image (multipart upload)

Plugin configuration

Image generation plugin

Add the Cloud APIM - Image generation backend plugin to your route:

{
  "enabled": true,
  "plugin": "cp:otoroshi_plugins.com.cloud.apim.otoroshi.extensions.aigateway.plugins.OpenAICompatImagesGen",
  "config": {
    "refs": ["image-model-entity-id"],
    "decode": false
  }
}

Parameter	Type	Default	Description
refs	array	[]	List of Image Model entity IDs
decode	boolean	false	When true, returns raw image bytes instead of JSON for single-image responses

Image editing plugin

Add the Cloud APIM - Image edition backend plugin to your route:

{
  "enabled": true,
  "plugin": "cp:otoroshi_plugins.com.cloud.apim.otoroshi.extensions.aigateway.plugins.OpenAICompatImagesEdit",
  "config": {
    "refs": ["image-model-entity-id"],
    "max_size_upload": 104857600
  }
}

Parameter	Type	Default	Description
refs	array	[]	List of Image Model entity IDs
max_size_upload	number	104857600 (100MB)	Maximum upload file size in bytes

Model routing

When multiple image model providers are configured in refs, you can target a specific provider using the model field in your request:

{
  "prompt": "a cat in space",
  "model": "providerName/modelName"
}

The provider can be referenced by:

• Entity name (slug): openai-provider/gpt-image-1
• Entity ID: image-model-id###gpt-image-1

If no provider prefix is specified, the first configured ref is used.

Entity configuration

An image model entity is configured with a provider type, connection settings, and generation/edition options:

{
  "id": "image-model-entity-id",
  "name": "My Image Provider",
  "description": "An image generation provider",
  "provider": "openai",
  "config": {
    "connection": {
      "base_url": "https://api.openai.com/v1",
      "token": "sk-xxx",
      "timeout": 180000
    },
    "options": {
      "generation": {
        "enabled": true,
        "model": "gpt-image-1",
        "size": "auto",
        "quality": "auto",
        "n": 1
      },
      "edition": {
        "enabled": true,
        "model": "gpt-image-1",
        "n": 1
      }
    }
  },
  "models": {
    "include": [],
    "exclude": []
  }
}

The token field supports comma-separated round-robin tokens — if the token contains commas, the system rotates through them.

Generation request

curl --request POST \
  --url http://myroute.oto.tools:8080/v1/images/generations \
  --header 'content-type: application/json' \
  --data '{
  "prompt": "a white siamese cat",
  "model": "gpt-image-1",
  "n": 1,
  "size": "1024x1024"
}'

Generation parameters

Parameter	Type	Description
prompt	string	The text prompt describing the image to generate
model	string	Model name (can include provider prefix)
n	integer	Number of images to generate
size	string	Image dimensions (e.g., 1024x1024, 1792x1024)
quality	string	Image quality (standard, hd, auto)
style	string	Image style (natural, vivid)
response_format	string	Response format (url or b64_json)
background	string	Background setting
moderation	string	Moderation level
output_compression	integer	Output compression level
output_format	string	Output format (e.g., png, jpeg)

Not all parameters are supported by all providers. See the models page for per-provider details.

Generation response

{
  "created": 1762441013,
  "data": [
    {
      "url": "https://...",
      "b64_json": null,
      "revised_prompt": "a white siamese cat sitting gracefully..."
    }
  ]
}

Image editing

See the image editing page for details on editing images with supported providers.