Llm Proxy

What This Skill Does

The LLM Proxy skill serves as a unified gateway for OpenClaw users to interact with over 22 different Large Language Model (LLM) providers. It acts as a middleware that sits between your request and the AI providers, offering centralized management, health monitoring, and enterprise-grade content safety auditing. Whether you are using local models via Ollama, open-source models through SiliconFlow, or commercial giants like OpenAI and Anthropic, this proxy standardizes the communication and provides a layer of security that ensures both input and output adhere to safety policies.

Installation

To add this skill to your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/394286006/llm-proxy

Once installed, you can manage the service directly through OpenClaw natural language commands like "启动llm-proxy" (start) or "llm-proxy状态" (status). For advanced users, script-based control is available in the scripts/ directory.

Use Cases

• Unified AI Development: Developers can switch between models (e.g., swapping DeepSeek for GPT-4) without changing application code, as the proxy standardizes the API format.
• Content Security Gatekeeping: Organizations can use the proxy's three-layer auditing system to block malicious instructions, SQL injections, or prohibited sensitive content before it reaches the model or the user.
• Real-time Monitoring: Track total request volume, error rates, and blocked content incidents via the health check endpoint, enabling better infrastructure observability.
• Cost Control & Management: Consolidate access keys and manage rate limits across various providers in one centralized configuration file.

Example Prompts

1. "启动llm-proxy并告诉我当前的运行状态。"
2. "llm-proxy状态，如果服务未响应请尝试重启。"
3. "停止llm-proxy，我需要修改配置文件中的代理端口。"

Tips & Limitations

• Performance: While the proxy supports streaming content security checks, be aware that complex keyword matching can add a few milliseconds of latency to your LLM interactions.
• Configuration: Always ensure your llm-proxy-config.json file is correctly formatted before restarting the service, as incorrect JSON syntax will prevent the proxy from starting.
• Security: The proxy is designed to listen on 127.0.0.1 by default. If you need to expose this to other machines on your network, ensure you implement appropriate firewall and authentication rules, as the default configuration favors local safety.
• Monitoring: Since there is no built-in automated monitoring daemon, integrate the /health endpoint with your existing tools like Prometheus or Uptime Kuma for reliable uptime tracking.