An intelligent model routing tool that switches between high-performance cloud LLMs and local Ollama instances based on task intent and user commands.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mlti-llm-fallback
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 mlti-llm-fallback 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 Multi-LLM Intelligent Switching?
Multi-LLM is a robust model management utility designed to optimize how developers interact with large language models. While the system defaults to using Claude Opus 4.5 for high-complexity tasks, it provides a specialized trigger to activate local model selection. This allows users of Openclaw Skills to leverage the power of local hardware for specific domains like coding, math, and translation, reducing reliance on external APIs and improving privacy.
By integrating this skill, you gain a dynamic execution environment that understands the context of your request. Whether you are performing heavy refactoring or simple text summarization, Multi-LLM ensures the right model is used for the right job, backed by a sophisticated fallback mechanism that guarantees reliability even if specific local models are missing.
Multi-LLM Intelligent Switching Use Cases
- Reducing API token consumption by offloading standard coding tasks to local models like Qwen2.5-Coder.
- Running privacy-sensitive reasoning and logic analysis locally using DeepSeek-R1.
- Speeding up translation and summarization workflows with lightweight, local Chinese language models.
- Maintaining workflow continuity through automated fallback chains when primary models are unavailable in the Openclaw Skills environment.
How Multi-LLM Intelligent Switching Works
- The system monitors user input for the specific trigger command
multi llm. - If the trigger is absent, the system routes the request to the default high-performance model, Claude Opus 4.5.
- When the trigger is detected, an intelligent detection logic scans the prompt for keywords related to coding, reasoning, or language.
- The skill maps the detected task type to a prioritized list of local Ollama models.
- If the primary local model (e.g., DeepSeek-R1 for reasoning) is not found, the system traverses a fallback chain to find the next best available local alternative.
Multi-LLM Intelligent Switching Setup
To get started with this skill in your Openclaw Skills setup, ensure Ollama is installed and the required models are pulled:
# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh
# Start the Ollama service
ollama serve
# Pull the recommended model set
ollama pull qwen2.5-coder:32b
ollama pull deepseek-r1:70b
ollama pull glm4:9b
ollama pull qwen3:32b
Verify your local environment with ollama list to ensure all models are ready for the switching logic.
Multi-LLM Intelligent Switching Data Schema & Taxonomy
The skill organizes model selection based on task categories and predefined model hierarchies:
| Task Category | Primary Model | Size | Fallback Priority |
|---|---|---|---|
| Coding | qwen2.5-coder:32b | 19GB | qwen2.5-coder:14b -> qwen3:32b |
| Reasoning | deepseek-r1:70b | 42GB | deepseek-r1:32b -> qwen3:32b |
| Chinese | glm4:9b | 5.5GB | qwen3:8b -> qwen3:32b |
| General | qwen3:32b | 20GB | qwen3:14b -> qwen3:8b |
Internal logic is handled via select-model.sh and fallback-demo.sh located in the scripts directory.
Multi-LLM Intelligent Switching Advanced Features
- Intelligent keyword detection supporting both English and Chinese technical terminology.
- Manual override commands such as
multi llm codingormulti llm reasoningto bypass automated detection. - Multi-tier fallback chains that prevent execution errors by cascading through available model versions.
- Programmatic access via shell scripts, allowing other Openclaw Skills to query the recommended model for a specific task string.
SKILL.md
Loading
Related Openclaw Skills
AutoFillIn
An intelligent browser automation skill for filling complex web forms and uploading files with persistent login support.
Polymarket FastLoop Trader
An automated trading skill for Polymarket fast markets driven by Binance price momentum and the Simmer API.
Bird Info Skill
A specialized research skill for AI agents to retrieve detailed bird species data, taxonomy, and conservation status from the Dongniao database.
Zeng Ming AI Strategy Assistant
An AI-powered knowledge assistant delivering strategic insights, business quotes, and industry observations from Professor Zeng Ming, former Dean of Hupan University.
Xiangji Translation Service
Xiangji Translation Service provides professional-grade text and image translation capabilities by integrating multiple global AI engines.
Better Auth Documentation Integration
A comprehensive technical reference skill for implementing modern, secure authentication in TypeScript projects using AI agents.
METADATA
Featured*
