smart-model-router
v1.0.0
Published
Intelligently route AI requests to optimal models based on task type, cost, speed, and quality
Maintainers
mind_workReadme
🧠 Smart Model Router
Intelligently route AI requests to the optimal model — Save 30%+ on costs while maintaining quality.
English
🚀 What is Smart Model Router?
Smart Model Router is an OpenClaw plugin that automatically selects the best AI model for your task based on:
- Task Type — Code generation? Use Claude Sonnet. Simple Q&A? Use GPT-4o-mini.
- Cost — Save up to 90% by routing to cheaper models when quality difference is minimal
- Speed — Get responses 2-3x faster by choosing low-latency models
- Quality — Ensure best output for critical tasks
No code changes needed — Just enable the plugin and let AI agents use it automatically.
✨ Key Features
- 🎯 Intelligent Routing — Automatically selects optimal model based on task type
- 💰 Cost Optimization — Reduce AI costs by 30-90% without sacrificing quality
- ⚡ Speed Priority — Route to fastest models for real-time interactions
- 🛡️ Auto Fallback — Automatically switch to backup model if primary fails
- 📊 Analytics — Track usage, costs, and performance metrics
- 🔧 Customizable — Define your own task-to-model mappings
- 🌐 Multi-Provider — Supports OpenAI, Anthropic, Google, DeepSeek, and more
📦 Installation
# Install the plugin
openclaw plugins install @openclaw/smart-model-router
# Or install from local directory
openclaw plugins install -l ./smart-router
# Restart OpenClaw Gateway
openclaw gateway restart⚙️ Configuration
Add to your OpenClaw config (~/.openclaw/config.yaml):
plugins:
entries:
smart-model-router:
enabled: true
config:
defaultStrategy: balanced # cost | speed | quality | balanced
enableLogging: true
enableFallback: true
models:
- name: gpt-4o
provider: openai
costPer1kTokens: 0.005
avgLatencyMs: 1500
qualityScore: 4.8
- name: gpt-4o-mini
provider: openai
costPer1kTokens: 0.0005
avgLatencyMs: 800
qualityScore: 4.2
- name: claude-sonnet-4
provider: anthropic
costPer1kTokens: 0.003
avgLatencyMs: 1200
qualityScore: 4.9🎮 Usage
As Agent Tool (Automatic)
AI agents will automatically call the smart_model_router tool:
User: Write a Python function to calculate Fibonacci numbers
AI Agent: [Calls smart_model_router tool]
{
"task_type": "code_generation",
"strategy": "balanced"
}
Tool Response:
{
"selected_model": "claude-sonnet-4",
"provider": "anthropic",
"reason": "Best overall balance for code_generation",
"estimated_cost_per_1k_tokens": 0.003,
"quality_score": 4.9
}
AI Agent: [Uses claude-sonnet-4 to generate code]CLI Commands
# View routing statistics
openclaw smart-router stats
# View recent routing logs
openclaw smart-router logs -n 50
# Test routing decision
openclaw smart-router test --task code_generation --strategy cost📊 Routing Strategies
| Strategy | Description | Best For | |----------|-------------|----------| | cost | Cheapest model | High-volume, non-critical tasks | | speed | Fastest response | Real-time chat, interactive apps | | quality | Best output | Critical tasks, complex problems | | balanced | Smart mix of all factors | General use (recommended) |
🎯 Task Types
| Task Type | Preferred Models | Use Case |
|-----------|------------------|----------|
| code_generation | Claude Sonnet, GPT-4o | Writing code |
| code_review | Claude Sonnet, GPT-4o | Reviewing code |
| qa | GPT-4o-mini, Claude Haiku | Simple questions |
| translation | GPT-4o-mini | Language translation |
| creative_writing | Claude Sonnet, GPT-4o | Stories, articles |
| summarization | GPT-4o-mini | Text summarization |
| data_analysis | GPT-4o, Claude Sonnet | Analyzing data |
| chat | GPT-4o-mini, Gemini Flash | Casual conversation |
💡 Examples
Example 1: Cost Optimization
Before: Using GPT-4o for all tasks → $0.005/1k tokens
After: Smart routing
- Simple Q&A → GPT-4o-mini ($0.0005/1k) — 90% savings
- Code generation → Claude Sonnet ($0.003/1k) — 40% savings
- Chat → Gemini Flash ($0.0002/1k) — 96% savings
Result: 30-70% overall cost reduction
Example 2: Speed Priority
Before: Using Claude Sonnet for chat → ~1200ms latency
After: Smart routing to Gemini Flash → ~500ms latency
Result: 2.4x faster responses
Example 3: Auto Fallback
Scenario: Claude Sonnet API is down
Before: Request fails, user sees error
After: Automatically switches to GPT-4o
Result: 99.9%+ uptime
🔧 Advanced Configuration
Custom Task Rules
plugins:
entries:
smart-model-router:
config:
taskRules:
my_custom_task:
- deepseek-coder
- gpt-4o
another_task:
- gemini-proAdd Custom Models
plugins:
entries:
smart-model-router:
config:
models:
- name: my-custom-model
provider: custom-provider
costPer1kTokens: 0.002
avgLatencyMs: 1000
qualityScore: 4.5
enabled: true📈 Analytics
View detailed statistics:
$ openclaw smart-router stats
📊 Smart Model Router Statistics
Total Calls: 1,234
Total Cost: $12.45
Avg Latency: 856ms
Success Rate: 99.8%
Model Usage:
gpt-4o-mini: 789 (63.9%)
claude-sonnet-4: 312 (25.3%)
gemini-flash: 133 (10.8%)
Strategy Usage:
balanced: 856 (69.4%)
cost: 234 (19.0%)
speed: 144 (11.6%)🤝 Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
- Fork the repository
- Create your feature branch (
git checkout -b feature/AmazingFeature) - Commit your changes (
git commit -m 'Add some AmazingFeature') - Push to the branch (
git push origin feature/AmazingFeature) - Open a Pull Request
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
🙏 Acknowledgments
📞 Support
中文
🚀 什么是智能模型路由？
智能模型路由是一个 OpenClaw 插件，可以根据以下因素自动为您的任务选择最佳 AI 模型：
- 任务类型 — 代码生成？使用 Claude Sonnet。简单问答？使用 GPT-4o-mini。
- 成本 — 在质量差异最小的情况下，通过路由到更便宜的模型节省高达 90% 的成本
- 速度 — 通过选择低延迟模型，获得 2-3 倍更快的响应
- 质量 — 确保关键任务的最佳输出
无需修改代码 — 只需启用插件，让 AI 代理自动使用它。
✨ 核心功能
- 🎯 智能路由 — 根据任务类型自动选择最优模型
- 💰 成本优化 — 在不牺牲质量的情况下降低 30-90% 的 AI 成本
- ⚡ 速度优先 — 为实时交互路由到最快的模型
- 🛡️ 自动回退 — 主模型失败时自动切换到备用模型
- 📊 分析统计 — 跟踪使用情况、成本和性能指标
- 🔧 可自定义 — 定义您自己的任务到模型映射
- 🌐 多提供商 — 支持 OpenAI、Anthropic、Google、DeepSeek 等
📦 安装
# 安装插件
openclaw plugins install @openclaw/smart-model-router
# 或从本地目录安装
openclaw plugins install -l ./smart-router
# 重启 OpenClaw Gateway
openclaw gateway restart⚙️ 配置
在 OpenClaw 配置文件中添加（~/.openclaw/config.yaml）：
plugins:
entries:
smart-model-router:
enabled: true
config:
defaultStrategy: balanced # cost | speed | quality | balanced
enableLogging: true
enableFallback: true
models:
- name: gpt-4o
provider: openai
costPer1kTokens: 0.005
avgLatencyMs: 1500
qualityScore: 4.8
- name: gpt-4o-mini
provider: openai
costPer1kTokens: 0.0005
avgLatencyMs: 800
qualityScore: 4.2🎮 使用方法
作为 Agent 工具（自动）
AI 代理会自动调用 smart_model_router 工具：
用户：写一个 Python 函数计算斐波那契数列
AI 代理：[调用 smart_model_router 工具]
{
"task_type": "code_generation",
"strategy": "balanced"
}
工具响应：
{
"selected_model": "claude-sonnet-4",
"provider": "anthropic",
"reason": "代码生成的最佳平衡选择",
"estimated_cost_per_1k_tokens": 0.003,
"quality_score": 4.9
}
AI 代理：[使用 claude-sonnet-4 生成代码]CLI 命令
# 查看路由统计
openclaw smart-router stats
# 查看最近的路由日志
openclaw smart-router logs -n 50
# 测试路由决策
openclaw smart-router test --task code_generation --strategy cost📊 路由策略
| 策略 | 描述 | 最适合 | |------|------|--------| | cost | 最便宜的模型 | 大量非关键任务 | | speed | 最快响应 | 实时聊天、交互式应用 | | quality | 最佳输出 | 关键任务、复杂问题 | | balanced | 所有因素的智能组合 | 通用场景（推荐） |
🎯 任务类型
| 任务类型 | 首选模型 | 使用场景 |
|---------|---------|---------|
| code_generation | Claude Sonnet, GPT-4o | 编写代码 |
| code_review | Claude Sonnet, GPT-4o | 代码审查 |
| qa | GPT-4o-mini, Claude Haiku | 简单问答 |
| translation | GPT-4o-mini | 语言翻译 |
| creative_writing | Claude Sonnet, GPT-4o | 故事、文章 |
| summarization | GPT-4o-mini | 文本摘要 |
| data_analysis | GPT-4o, Claude Sonnet | 数据分析 |
| chat | GPT-4o-mini, Gemini Flash | 日常对话 |
💡 示例
示例 1：成本优化
之前： 所有任务使用 GPT-4o → $0.005/1k tokens
之后： 智能路由
- 简单问答 → GPT-4o-mini ($0.0005/1k) — 节省 90%
- 代码生成 → Claude Sonnet ($0.003/1k) — 节省 40%
- 聊天 → Gemini Flash ($0.0002/1k) — 节省 96%
结果： 总体成本降低 30-70%
示例 2：速度优先
之前： 聊天使用 Claude Sonnet → ~1200ms 延迟
之后： 智能路由到 Gemini Flash → ~500ms 延迟
结果： 响应速度提升 2.4 倍
示例 3：自动回退
场景： Claude Sonnet API 宕机
之前： 请求失败，用户看到错误
之后： 自动切换到 GPT-4o
结果： 99.9%+ 正常运行时间
📈 分析统计
查看详细统计：
$ openclaw smart-router stats
📊 智能模型路由统计
总调用次数：1,234
总成本：$12.45
平均延迟：856ms
成功率：99.8%
模型使用：
gpt-4o-mini: 789 (63.9%)
claude-sonnet-4: 312 (25.3%)
gemini-flash: 133 (10.8%)
策略使用：
balanced: 856 (69.4%)
cost: 234 (19.0%)
speed: 144 (11.6%)🤝 贡献
欢迎贡献！请随时提交 Pull Request。
- Fork 仓库
- 创建功能分支 (
git checkout -b feature/AmazingFeature) - 提交更改 (
git commit -m 'Add some AmazingFeature') - 推送到分支 (
git push origin feature/AmazingFeature) - 打开 Pull Request
📄 许可证
本项目采用 MIT 许可证 - 详见 LICENSE 文件。
🙏 致谢
📞 支持
- 📖 文档
- 💬 Discord 社区
- 🐛 问题追踪
⭐ If you find this plugin useful, please give it a star!
⭐ 如果您觉得这个插件有用，请给它一个星标！