smart-session-routing

🧠 Smart Session Routing

中文文档

An intelligent session routing skill for AI chat applications — helps AI agents decide whether a user's message should continue the current conversation or start a new one.

✨ Features

• 🎯 5 Decision Signals — Semantic relevance, explicit intent, context window, time gap, conversation health
• 🔄 3 Implementation Approaches — AI self-judgment (prompt-based), backend middleware, frontend heuristics
• 🌐 Bilingual Support — Chinese & English intent detection out of the box
• 📊 Complete Decision Matrix — Combined signal evaluation with edge case handling
• 🧩 Framework Agnostic — Works with any AI model, any frontend framework, any backend stack

📦 Installation

Option 1: Direct Copy

git clone https://github.com/YOUR_USERNAME/smart-session-routing.git
cp -r smart-session-routing /path/to/your-project/.agent/skills/

Option 2: Git Submodule (Recommended)

cd /path/to/your-project
git submodule add https://github.com/YOUR_USERNAME/smart-session-routing.git .agent/skills/smart-session-routing

Update to latest version:

git submodule update --remote .agent/skills/smart-session-routing

📂 Structure

smart-session-routing/
├── SKILL.md                           # Core guide (~750 words, always loaded)
├── references/
│   ├── decision-framework.md          # Detailed scoring model & thresholds
│   └── implementation-patterns.md     # Architecture patterns & design principles
└── examples/
    ├── conversation-scenarios.md      # 14 real-world routing scenarios
    └── session-router-prompt.md       # Prompt template for AI self-judgment

🚀 Quick Start

How It Works

When a user sends a message, the router evaluates 5 signals:

Incoming Message
    │
    ├─ Explicit new-session intent?  ──YES──→ 🆕 NEW SESSION
    │
    ├─ Context window >80% full?     ──YES──→ 🆕 NEW SESSION (with summary)
    │
    ├─ Semantic similarity < 0.3?    ──YES──→ ❓ PROMPT USER
    │
    ├─ Time gap > threshold?         ──YES──→ ❓ PROMPT USER
    │
    └─ None of the above             ──────→ ✅ CONTINUE

Approach 1: AI Self-Judgment (Fastest to Integrate)

Add the prompt template from examples/session-router-prompt.md to your system prompt. The AI will output a routing decision as a JSON comment before each response:

<!--session_route: {"decision": "continue", "confidence": 0.9}-->

Approach 2: Lightweight Middleware

Implement a routing service that intercepts messages before they reach the AI model. See references/implementation-patterns.md for three architecture patterns.

Approach 3: Hybrid (Production Grade)

Combine client-side heuristics (instant intent/time checks) with server-side intelligence (semantic analysis). Best for large-scale applications.

See examples/conversation-scenarios.md for 14 real-world scenarios to calibrate your routing logic.

⚙️ Configuration

Key thresholds to tune for your use case:

🤝 Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'feat: add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

MIT