How to Run an AI Image Generation Agent with OE Runtime

Turn text descriptions into high-quality images automatically. OE Runtime lets you run an AI image generation agent locally — no cloud platform, no monthly subscription, just a YAML file and a binary.

What You Need

• Node.js 18+ installed, or the OE Runtime binary from openenthrium.com/runtime
• An API key from OpenAI, Anthropic, or any supported LLM provider
• A Replicate API key for image generation
• A text editor — nothing else

Create the Project Folder

Create a folder called image-generation/ and add these two files:

The Skill Files

Create SKILL.md inside an image-generation/ directory. This is the portable skill — it follows the agentskills.io format and runs unchanged on Claude, Cursor, Windsurf, or OE Runtime:

---
name: image-generation
description: Generate images from text descriptions using AI
license: Apache-2.0
metadata:
  author: Open Enthrium
  version: "1.0"
---

You are a creative image generation agent. Craft detailed prompts and generate high-quality images.
If the concept is vague, enhance it with artistic style, lighting, and composition details before generating.
Complete all steps fully before writing your report.

## Step 1: Plan the Prompt
Design a detailed image generation prompt for a professional product showcase photo.
The subject: a sleek laptop on a minimalist desk with soft natural lighting.
Enhance with: camera angle, lighting style, color palette, mood, and aspect ratio (16:9).

## Step 2: Generate Image
Submit the prompt to the OpenAI Image connector.
POST /images/generations with body:
{
  "model": "gpt-image-1",
  "prompt": "<detailed prompt from previous step>",
  "n": 1,
  "size": "1024x1024"
}
The response will contain data[0].local_path with the saved image file path. Report that path.

## Step 3: Report
Summarize the result:
- The final prompt used
- Any enhancements made to the original concept
- Image URL
- Style and composition notes for future reference

Create agent.yaml in the same directory to wire the skill to your connector:

name: Image Creator
description: Generate images from text descriptions using AI
connectors:
  - connection_name: OpenAI Image
    connection_type: openai-image
skills:
  - path: ./
    trigger_type: auto

The Config File

Create oe-config.json in the same directory:

{
  "llm": {
    "provider": "openai",
    "model": "gpt-4o",
    "apiKey": "YOUR_OPENAI_API_KEY"
  },
  "server": {
    "enabled": false,
    "port": 3333,
    "apiKey": "your-secret-api-key"
  },
  "connectors": [
    {
      "connection_name": "OpenAI Image",
      "connection_type": "openai-image",
      "bearerToken": "YOUR_OPENAI_API_KEY",
      "baseUrl": "https://api.openai.com/v1",
      "model": "gpt-image-1"
    }
  ]
}

Replace YOUR_OPENAI_API_KEY with your key from platform.openai.com/api-keys.

Download OE Runtime

Run the Agent

From the parent folder containing your skill directory:

1 npx recommended

npx -y @openenthrium/oe-runtime@latest ./image-generation

2 Windows

oe-runtime-win.exe ./image-generation

3 macOS

chmod +x oe-runtime-macos
./oe-runtime-macos ./image-generation

First run blocked? System Settings → Privacy & Security → Allow Anyway.

4 Linux

chmod +x oe-runtime-linux
./oe-runtime-linux ./image-generation

5 API Server integration

Add a "server" block to oe-config.json, then start with --serve:

{
  "llm": { ... },
  "server": { "enabled": true, "port": 3333, "apiKey": "your-secret-key" },
  "connectors": [ ... ]
}

npx -y @openenthrium/oe-runtime@latest --serve --config oe-config.json

Run with inline YAML:

curl -X POST http://localhost:3333/run \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your-secret-key" \
  -d '{"yaml": "...", "params": {}}'

Or run from a file on the server:

curl -X POST http://localhost:3333/run-file \
  -H "Content-Type: application/json" \
  -H "X-API-Key: your-secret-key" \
  -d '{"file": "/path/to/agent.yaml", "params": {}}'

Use Cases

• Product photography — Generate studio-quality product shots from written descriptions without booking a photographer or renting equipment. Describe the product, background, lighting style, and angle, and the agent produces a professional-grade image ready for web or print.
• Marketing & advertising creative — Produce social media visuals, ad banners, email headers, and campaign imagery at scale. Brief the agent with your brand colours, tone, and message, and get multiple creative variations in minutes instead of days.
• UI & web design mockups — Create realistic placeholder images, hero artwork, icon sets, and illustration packs for design prototypes. Stop using stock photos that look generic — generate visuals that match your exact layout and aesthetic.
• Brand content at scale — Maintain visual consistency across hundreds of assets by embedding your brand guidelines directly in the prompt template. The agent applies the same colour palette, style, and composition rules every time it runs.
• E-commerce catalogue imagery — Automate product image creation for large catalogues where individual photography is impractical. Feed the agent a product list and generate a consistent image set for every SKU, ready for your storefront.
• Game & app concept art — Rapidly prototype characters, environments, UI elements, and storyboards for games or mobile apps. Use the agent to explore multiple visual directions before committing design resources to a single style.
• Real estate & interior visualisation — Generate photorealistic renders of rooms, facades, or renovation concepts from a text brief. Useful for architects, agents, and interior designers who need quick visual proposals for clients.
• Training data generation — Create large volumes of labelled synthetic images for machine learning datasets. Control composition, lighting, and subject variation through prompt engineering to produce diverse, high-quality training sets.