Import custom safetensors models, including LoRAs and checkpoints, into Runware to perform server-side image inference using unique AIR handles.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install bring-your-own-model
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 bring-your-own-model 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 Bring Your Own Model?
The bring-your-own-model workflow enables developers to transition custom-trained weights directly into a cloud-hosted execution environment. By uploading your own LoRA, checkpoint, lycoris, vae, or embedding files in safetensors format to Runware, you create a first-class catalog model with a permanent AIR handle. Once registered, these models run within standard image generation tasks, avoiding the need to manage infrastructure for custom model hosting.
This feature provides a key asset within the ecosystem of Openclaw Skills, allowing agents and developers to programmatically scale their personalized generative assets. Using this integration, custom models perform with the same efficiency as native catalog models. Developers can manage their proprietary assets while optimizing image generation pipelines.
Bring Your Own Model Use Cases
- Hosting a custom-trained SDXL or FLUX LoRA to generate consistent brand assets or specific artistic styles.
- Uploading a specialized checkpoint file to run high-throughput API inferences without managing dedicated GPU infrastructure.
- Programmatically importing and organizing fine-tuned weights inside custom developer agent workflows leveraging Openclaw Skills.
- Managing model lifecycles (import, update, delete) dynamically within automated generative media platforms.
How Bring Your Own Model Works
- Query the Schema: Resolve the modelUpload or modelManagement schema using live tools to verify the supported architectures and required payload attributes.
- Configure Model Identifiers: Define a unique AIR (e.g., runware:id@n) alongside standard metadata including name, category, and base architecture.
- Initiate Upload: Submit the upload request specifying the public downloadURL of the safetensors file.
- Verify Readiness: Query model availability directly or run a test inference to check if the asset is active, bypassing stalling task status fields.
- Execute Inference: Call the custom model directly in generation requests via its newly registered AIR.
Bring Your Own Model Setup
To integrate the bring-your-own-model capability into your environment, prepare your tool configuration and ensure the proper SDK or CLI options are available. Ensure your workspace includes the Openclaw Skills runtime environment.
First, check available schemas and model status:
# Verify active model configurations and architectures
runware-run --schema modelUpload
# Query current available custom models
runware-models --list
Ensure that your workspace contains valid Runware API keys to authorize model management requests.
Bring Your Own Model Data Schema & Taxonomy
The skill processes specific parameters to register and identify your custom safetensors model.
| Parameter | Type | Required | Description |
|---|---|---|---|
downloadURL |
String | Yes | Accessible HTTP/HTTPS URL pointing to the model file |
category |
String (Enum) | Yes | Category of model: lora, checkpoint, lycoris, vae, embeddings |
architecture |
String | Yes | Base family identifier, matching valid strings for FLUX or SDXL |
air |
String | Yes | Permanent handle formatted as runware:<id>@<n> |
uniqueIdentifier |
String | Yes | Used alongside AIR for idempotency and lifecycle operations |
format |
String | Yes | Must be set to safetensors |
private |
Boolean | No | Determines model visibility, defaults to true |
positiveTriggerWords |
String | No | Recommended trigger phrases for LoRA activations |
Bring Your Own Model Advanced Features
- Idempotent Registering: Re-submitting an identical AIR and unique identifier checks status and returns the existing model instead of creating duplicate instances.
- Modern Lifecycle Control: Supports the successor
modelManagementtool, enabling swiftimport,update, and synchronousdeleteoperations by AIR parameters. - Robust Readiness Validation: Avoids stuck task state issues by directly checking model listings or issuing lightweight test inferences.
- Dynamic Parameter Mapping: Passes default parameters such as
defaultWeight,defaultSteps, anddefaultCFGstraight to inference tasks automatically, streamlining configuration within broader Openclaw Skills applications.
SKILL.md
Loading
Related Openclaw Skills
Character Consistency
Openclaw Skills for character consistency generates new AI images that keep the same subject recognizable across scenes, poses, outfits, and styles.
Learning English from TV Series (DramaLex)
DramaLex turns a TV episode or film into a complete English-learning loop with subtitle mining, CEFR calibration, drills, and review.
Cinematic Video
Generate short, shot-directed videos that feel filmed rather than synthesized, using precise cinematic prompts and live model-aware workflows.
Composite Scene
Composite Scene merges multiple real images into one coherent AI-generated scene by reconciling placement, scale, lighting, and perspective.
Agent Reputation
Openclaw Skills for Agent Reputation helps you discover, vet, register, claim, rate, and contact autonomous AI agents with provenance-aware trust and consent-first controls.
Lovense LAN Control
Openclaw Skills for Lovense LAN Control turns a zero-dependency Python CLI into safe, local-network control for Lovense toys and feedback scenes.
Featured*
