# Multi-Provider Agent Model Assignment System > **💼 Developer Note**: This SKILL was developed entirely by **Leo, the OpenClaw AI Assistant** **Version**: 1.0.0 **Created**: 2026-03-04 **Developer**: Leo (OpenClaw AI) **Development**: AI autonomous development, testing, and documentation --- ## 🎯 Overview The Multi-Provider Agent Model Assignment System is a powerful LLM management tool supporting: - **Multi-Cloud Providers** - Alibaba Cloud, OpenAI, Anthropic, Zhipu AI, Baidu, etc. - **Local Deployments** - Ollama, LM Studio, vLLM local model services - **Auto-Discovery** - Automatically scan and discover local model services - **Zero-Code Configuration** - Pure YAML config, no code changes needed - **Smart Routing** - Auto-select optimal model based on task type - **Cost Optimization** - Auto-select most cost-effective model combination --- ## 🚀 Quick Start ### 1. Install Dependencies ```bash pip install pyyaml requests ``` ### 2. List Available Models ```bash python skills/multi-provider-agents/scripts/provider_manager.py list-models ``` ### 3. Scan Local Models ```bash python skills/multi-provider-agents/scripts/provider_manager.py scan ``` ### 4. Configure Providers Edit `skills/multi-provider-agents/config/models.yaml` to add provider configurations. --- ## 📋 Features ### Multi-Provider Support Support unlimited cloud providers, just add to config: ```yaml providers: openai: name: OpenAI type: openai-compatible enabled: true base_url: https://api.openai.com/v1 api_key: sk-xxx models: - id: openai/gpt-4 name: GPT-4 context_window: 128000 ``` ### Local Model Support Support multiple local model services: #### Ollama ```yaml ollama: name: Ollama (Local) type: openai-compatible enabled: true base_url: http://localhost:11434/v1 models: - id: ollama/llama3 name: Llama 3 local: true ``` #### LM Studio ```yaml lmstudio: name: LM Studio (Local) type: openai-compatible enabled: true base_url: http://localhost:1234/v1 models: - id: lmstudio/local-model name: Local Model local: true ``` #### vLLM ```yaml vllm: name: vLLM (Local) type: openai-compatible enabled: true base_url: http://localhost:8000/v1 models: - id: vllm/mistral-7b name: Mistral 7B local: true ``` ### Auto-Discovery System automatically scans local model services: ```yaml local_discovery: enabled: true scan_interval: 300 # Scan interval (seconds) endpoints: - name: Ollama url: http://localhost:11434/api/tags type: ollama - name: LM Studio url: http://localhost:1234/v1/models type: openai-compatible ``` ### Smart Routing Auto-select optimal model by task type: ```yaml routing: task_routing: coding: bailian/qwen3-coder-plus # Coding tasks reasoning: bailian/qwen3-max # Reasoning tasks long_text: bailian/kimi-k2.5 # Long text agent: bailian/glm-5 # Agent orchestration daily: bailian/qwen3.5-plus # Daily tasks ``` ### Agent Model Assignment Configure dedicated models for each agent: ```yaml agents: Agent_Coordinator: # ⚠️ Replace with your agent name models: primary: bailian/glm-5 fallback: bailian/qwen3.5-plus preferences: cost_optimize: false latency_optimize: true Agent_Frontend_Dev: # ⚠️ Replace with your agent name models: primary: bailian/qwen3-coder-plus fallback: ollama/codellama preferences: code_quality: true ``` --- ## 🔧 Management Commands ### List All Providers ```bash python skills/multi-provider-agents/scripts/provider_manager.py list-providers ``` ### List All Models ```bash python skills/multi-provider-agents/scripts/provider_manager.py list-models ``` ### Scan Local Models ```bash python skills/multi-provider-agents/scripts/provider_manager.py scan ``` ### Add New Provider ```bash python skills/multi-provider-agents/scripts/provider_manager.py add ``` ### Enable/Disable Provider ```bash # Enable python skills/multi-provider-agents/scripts/provider_manager.py enable # Disable python skills/multi-provider-agents/scripts/provider_manager.py disable ``` --- ## 📁 Directory Structure ``` skills/multi-provider-agents/ ├── SKILL.md # Skill description ├── scripts/ │ ├── provider_manager.py # Provider manager │ └── model_router_v2.py # Model router ├── config/ │ └── models.yaml # Model configuration ├── docs/ │ ├── README_zh.md # Chinese documentation │ └── README_en.md # English documentation └── examples/ ├── add_openai.yaml # OpenAI config example ├── configure_ollama.yaml # Ollama config example └── agent_assignment.yaml # Agent assignment example ``` --- ## 💡 Use Cases ### Use Case 1: Add OpenAI 1. Edit `config/models.yaml` 2. Uncomment OpenAI configuration block 3. Fill in API Key 4. Run `python scripts/provider_manager.py list-models` to verify ### Use Case 2: Configure Local Ollama 1. Install Ollama: `https://ollama.ai` 2. Pull model: `ollama pull llama3` 3. Edit config to add Ollama block 4. Run `python scripts/provider_manager.py scan` to scan ### Use Case 3: Agent Uses Local Model Edit `agents` section in `config/models.yaml`: ```yaml agents: Agent_Frontend_Dev: # ⚠️ Replace with your agent name models: primary: bailian/qwen3-coder-plus # Cloud fallback: ollama/codellama # Local fallback preferences: local_first: true # Prefer local ``` --- ## 📊 Cost Optimization System supports cost tracking and optimization: ```yaml cost_tracking: enabled: true currency: CNY budget_monthly: 100.0 # Monthly budget alert_threshold: 0.8 # Alert threshold (80%) routing: cost_optimize: true # Enable cost optimization ``` --- ## 🔒 Security Recommendations 1. **Don't commit API Keys to Git** ```bash # Use environment variables api_key_env: OPENAI_API_KEY # Or add to .gitignore echo "config/models.yaml" >> .gitignore ``` 2. **Rotate API Keys Regularly** ```bash python scripts/provider_manager.py rotate-key ``` 3. **Use Local Models to Reduce Costs** ```yaml routing: local_first: true ``` --- ## 🤝 Contributing Issues and Pull Requests are welcome! --- ## 📄 License MIT License --- **Last Updated**: 2026-03-04 **Version**: 3.0.0