An intelligent, zero-friction model routing engine that detects task intent—such as vision, coding, or complex reasoning—and automatically switches to the optimal AI model.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install smart-model-switcher-v4
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 smart-model-switcher-v4 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 Smart Model Switcher V4?
Smart Model Switcher V4 is a sophisticated multi-modal perception layer designed for AI agents within the Openclaw Skills ecosystem. It eliminates the need for manual model selection by analyzing incoming messages for images, code snippets, or reasoning triggers in real-time. By dynamically routing requests to specialized models like qwen3-vl-plus for vision or glm-5 for development, it ensures high-performance output while optimizing resource usage.
The V4 update introduces advanced context analysis, allowing the agent to distinguish between pure image recognition and complex visual reasoning. This skill operates at a session level, meaning it can maintain different model configurations across multiple communication channels—such as Telegram, WebChat, or Feishu—without interference.
Smart Model Switcher V4 Use Cases
- Automatically switching to a vision model when a user uploads an error screenshot or diagram.
- Routing complex programming requests to high-performance coding models like glm-5 or qwen-coder-plus.
- Triggering deep reasoning models for mathematical proofs or logical analysis when specific keywords are detected.
- Handling long-form document analysis using models with 1M+ context windows like qwen3.5-plus.
- Cost-optimization by defaulting to turbo models for simple conversational tasks.
How Smart Model Switcher V4 Works
- The skill intercepts the incoming user message to analyze content types (images, attachments) and keywords.
- It matches the analyzed intent against a predefined decision tree (Vision > Code > Reasoning > General).
- The system checks model availability, API key status, and plan constraints.
- It executes a seamless session-level switch using the session_status tool without requiring user intervention.
- The AI agent provides a brief notification to the user about the switch and proceeds to execute the task with the optimized model.
Smart Model Switcher V4 Setup
To deploy Smart Model Switcher V4 within your environment, configure your model preferences in your agent settings:
# Example configuration for model preferences
{
"modelPreferences": {
"vision": "bailian/qwen3-vl-plus",
"coding": "bailian/glm-5",
"reasoning": "bailian/qwq-plus",
"general": "bailian/qwen3.5-plus",
"fast": "bailian/qwen-turbo"
}
}
Ensure that the corresponding API keys for Bailian or GLM providers are active in your Openclaw Skills dashboard.
Smart Model Switcher V4 Data Schema & Taxonomy
The skill manages model routing data through a hierarchical preference mapping and session-state tracking.
| Data Category | Field | Description |
|---|---|---|
| Model Preferences | vision, coding, reasoning, fast | Defines the primary model ID for specific task types. |
| Intent Detection | keywords, file_ext | Maps strings like '.py' or 'debug' to specific model categories. |
| Session Status | sessionKey, currentModel | Tracks the active model for each independent communication channel. |
| Fallback Logic | fallbackModel | Specifies the secondary model if the primary choice is unavailable. |
Smart Model Switcher V4 Advanced Features
- Multi-channel Independence: Maintain separate active models for WebChat, Telegram, and Feishu simultaneously.
- Visual Reasoning Logic: Automatically identifies if an image requires OCR, chart analysis, or visual coding help.
- Zero-Latency Switching: Uses session-level status updates to change models mid-conversation without dropping context.
- Custom Keyword Mapping: Ability to extend the detection engine with domain-specific triggers for specialized workflows.
SKILL.md
Loading
Related Openclaw Skills
Tavily Search Skill
A powerful web search integration that delivers LLM-optimized snippets, content scores, and metadata for precise AI research.
Hike Route Planner
A professional-grade tool for generating detailed hiking itineraries, safety briefs, and logistical plans for outdoor leaders and enthusiasts.
ShieldBot Security
An autonomous transaction firewall and security auditor for BNB Chain and major EVM networks.
Polymarket Weather Scanner
A professional-grade meteorological arbitrage tool that identifies discrepancies between Polymarket temperature betting lines and high-accuracy weather models.
Netease Cloud Music Hot Trend
An automated data analysis skill for tracking public music trends, playlist performance, and artist metrics on Netease Cloud Music.
Skill Creator
A specialized toolset for building, refining, and auditing modular AI agent skills that extend model capabilities with specialized knowledge.
Featured*
