The mflux image generation router provides a distributed, MLX-native framework for routing image requests across a fleet of Apple Silicon devices.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mflux-image-router
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install mflux-image-router using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is mflux Image Generation Router?
The mflux image generation router is a specialized skill designed to optimize local AI workflows by distributing heavy image synthesis tasks across multiple Mac devices. By utilizing MLX-native frameworks, it enables high-speed image generation on Apple Silicon without the overhead of cloud subscriptions or API fees. This skill is a core part of the Openclaw Skills ecosystem, allowing developers to treat a local fleet of Mac Minis, Studios, or MacBooks as a single, powerful image generation engine.
Rather than locking up a single machine's GPU for 20 seconds per image, this router intelligently identifies the least-loaded node in your network to process requests. It supports industry-leading models like Flux Dev and Flux Schnell, ensuring that high-quality visual content creation can happen concurrently with LLM inference or other developer tasks without performance degradation.
mflux Image Generation Router Use Cases
- Scaling local content creation pipelines without incurring DALL-E or Midjourney costs.
- Distributed image generation for multi-agent systems where one agent handles logic and another handles visualization.
- Optimizing GPU memory usage on Apple Silicon by offloading image tasks to idle Mac fleet nodes.
- High-volume image iteration for design prototyping using the ultra-fast z-image-turbo model.
How mflux Image Generation Router Works
- The user or agent sends a POST request to the central herd router with prompt and model parameters.
- The router queries the fleet health status to check which nodes have the required mflux model loaded and available GPU memory.
- The system selects the optimal node based on current CPU load and latency metrics.
- The chosen node executes the mflux generation using the MLX framework optimized for Apple Silicon.
- The generated PNG bytes are routed back through the API to the requester, accompanied by headers indicating which node performed the work.
mflux Image Generation Router Setup
To integrate this into your Openclaw Skills workflow, follow these installation steps:
# Install the fleet manager
pip install ollama-herd
# Start the central router
herd
# Start a node on each Apple Silicon device
herd-node
# Ensure mflux is available on the worker nodes
uv tool install mflux
# Enable image generation via the API
curl -X POST http://localhost:11435/dashboard/api/settings \
-H "Content-Type: application/json" \
-d '{"image_generation": true}'
mflux Image Generation Router Data Schema & Taxonomy
The skill manages configuration and performance data through a centralized directory structure:
| Data Type | Path / Format | Description |
|---|---|---|
| Latency DB | ~/.fleet-manager/latency.db |
SQLite database tracking node response times |
| Logs | ~/.fleet-manager/logs/herd.jsonl |
JSONL logs of all image generation requests |
| Output | image/png |
Raw binary PNG data returned via API |
| Metadata | X-Fleet-Node Header |
Identifies the specific hardware used for the task |
mflux Image Generation Router Advanced Features
- Multi-model support including z-image-turbo for speed and flux-dev for maximum photorealism.
- Intelligent queue management that prevents GPU memory contention with LLM inference tasks.
- Metadata tagging for per-project tracking and cost-savings analysis within the dashboard.
- Automated quantization support (3-8 bit) to balance image quality with available hardware VRAM.
- Real-time fleet monitoring dashboard accessible at port 11435 with specific [IMAGE] badges for active generation tasks.
SKILL.md
Loading
Related Openclaw Skills
Enterprise Change Management Lead
An expert change management advisor designed to diagnose adoption risks and drive stakeholder readiness for complex enterprise transformations.
Local Speech-to-Text Transcription
A secure, fleet-distributed local speech-to-text transcription skill powered by Qwen ASR and MLX for Apple Silicon.
Fleet Embeddings
Distributed embedding generation across a device fleet for high-performance RAG and semantic search pipelines.
Coffee Shop Video Maker
A specialized AI video generation tool designed to transform the daily operations of independent cafes and espresso bars into high-converting social media content.
ClawGuard
ClawGuard is a specialized security assistant that interprets scan reports and guides the remediation of configuration vulnerabilities within OpenClaw.
Writing Coach Pro
A sophisticated AI writing mentor that tracks your style preferences and provides multi-pass analysis to improve document quality and consistency.
Featured*
