Arya Model Router

What This Skill Does

The Arya Model Router is a sophisticated intelligent traffic controller for your OpenClaw agent ecosystem. Its primary function is to optimize token usage and API expenditures by dynamically selecting the most cost-effective Large Language Model (LLM) for any given request. Instead of defaulting to high-cost, high-latency models for mundane chat or simple queries, the router analyzes the intent and complexity of your prompt to choose between three tiers: 'cheap', 'default', or 'pro'. Beyond simple selection, the skill introduces an architectural shift where it creates a concise brief or compresses context before routing complex tasks to specialized sub-agents, ensuring that high-performance models only receive the most relevant data. By offloading heavy lifting to specific sub-agents and keeping the main conversational agent on an economical tier, users can drastically reduce their monthly AI bill without sacrificing quality for demanding tasks.

Installation

To integrate the Arya Model Router into your OpenClaw environment, ensure you have both bash and python3 installed on your system. Run the following command in your terminal to fetch the repository and register the skill:

clawhub install openclaw/skills/skills/staratheris/arya-model-router

Once installed, verify the configuration files (router.py and rules.json) in your skills directory. You may customize the specific model mappings in the rules.json file if you wish to override the default provider endpoints.

Use Cases

• Automated Cost Control: Automatically handle daily interactions and trivial follow-ups with budget-friendly models.
• Complex Data Processing: Automatically escalate tasks involving large codebases or complex data structures to the 'pro' model tier once a condensed brief is generated.
• Context Optimization: Use the briefing feature to summarize long chat histories before feeding them into an expensive model, saving thousands of tokens per request.

Example Prompts

1. "Router: respond to this simple greeting in cheap mode to keep costs minimal."
2. "Router: analyze this entire project repository and summarize the architectural bottlenecks."
3. "Router: draft a technical report based on the provided logs, use the pro model."

Tips & Limitations

• Tuning: Regularly inspect the rules.json file to refine the heuristics used by the classifier.
• Latency: Be aware that routing decisions introduce a negligible, yet present, delay in the initial response time.
• Sub-agents: Ensure your sub-agents have the necessary permissions configured in OpenClaw, as the router relies on their availability for 'pro' tasks.
• Precision: While the router is highly effective, very ambiguous prompts may occasionally result in model selection that doesn't perfectly match your expectations; manual overrides are recommended for critical workflows.