smart-model-switcher-v2

# 🧠 Smart Model Switcher V2 (Optimized)

**Zero-Latency • No Restart • Runtime Switching**

## 🎯 What's New in V2

| Feature | V1 | V2 |
|---------|----|----|
| **Restart Required** | ❌ Yes | ✅ No |
| **Switch Latency** | 5-10s | <100ms |
| **Model Preloading** | ❌ No | ✅ Yes |
| **Parallel Processing** | ❌ No | ✅ Yes |
| **Auto Model Discovery** | ❌ No | ✅ Yes |
| **Fallback Logic** | Basic | Advanced |
| **Performance** | Low | High |

## 🚀 New Features

### 1. Zero-Latency Switching
- No gateway restart needed
- Runtime model selection
- <100ms switching latency
- User-transparent operation

### 2. Model Preloading
- All plan models preloaded at startup
- Instant switching between models
- No API connection delays
- Connection pooling

### 3. Intelligent Task Classification
- Multi-keyword detection
- Context-aware analysis
- Confidence scoring
- Fallback to default if uncertain

### 4. Parallel Model Processing
- Multiple models ready simultaneously
- Fast failover if model unavailable
- Load balancing across models
- Automatic retry logic

### 5. Auto Model Discovery
- Detects new models in your plan
- Auto-updates model registry
- No manual configuration needed
- Real-time plan synchronization

### 6. Advanced Fallback
- Multi-tier fallback chain
- Graceful degradation
- User notification on fallback
- Logs all fallback events

## 📊 Model Selection Matrix (Optimized)

| Task Type | Primary Model | Fallback 1 | Fallback 2 | Latency |
|-----------|--------------|------------|------------|---------|
| **写小说/创意写作** | qwen3.5-plus | qwen3.5-397b | qwen-plus | <50ms |
| **写代码/编程** | qwen3-coder-plus | qwen3-coder-next | qwen3.5-plus | <50ms |
| **复杂推理/数学** | qwen3-max | qwen3.5-plus | qwen-plus | <50ms |
| **数据分析** | qwen3.5-plus | qwen3-max | qwen-plus | <50ms |
| **日常对话** | qwen3.5-plus | qwen-plus | qwen-turbo | <30ms |
| **长文档处理** | qwen3.5-plus | qwen3.5-397b | qwen-plus | <50ms |
| **Debug/修复** | qwen3-coder-plus |...