semantic-router

What This Skill Does

Semantic Router v7.9.3 is a production-grade session routing system for OpenClaw, designed to intelligently dispatch user requests to the most optimal AI model pools. By utilizing a four-layer recognition process—system filtering, keyword analysis, instruction detection, and semantic similarity scoring—the skill ensures that every query is handled by the model best suited for the specific task. It addresses critical issues such as session resets caused by background Cron jobs and the fragmentation of long-running workflows. With a sophisticated four-pool architecture (Highspeed, Intelligence, Humanities, and Agentic) and a five-branch routing logic, this tool maintains session context while optimizing for cost, speed, and precision.

Installation

Installation is streamlined through ClawHub. First, execute clawhub install https://clawhub.ai/halfmoon82/semantic-router to pull the latest repository. After installation, navigate to the skill directory and run the setup wizard with python3 ~/.openclaw/workspace/skills/semantic-router/scripts/setup_wizard.py. This script will analyze your existing configuration, scan available models, and generate the necessary pools.json and tasks.json files. Finally, ensure any active Cron jobs are isolated from your main session keys to prevent interference during task switching.

Use Cases

This skill is perfect for power users who juggle multiple AI workflows simultaneously. It excels in:

• Separating high-frequency administrative tasks from deep development coding sessions.
• Managing complex multi-step agentic workflows that require massive context windows.
• Automatically shifting to faster, cost-efficient models for simple information retrieval.
• Preserving professional writing contexts while switching models for specialized research tasks.

Example Prompts

1. "帮我写个 Python 爬虫脚本，用来抓取公开的实时天气数据。"
2. "处理并整理这份长文档，将其中的关键论点总结成 PPT 大纲。"
3. "继续刚才的逻辑，但是请用更严谨的学术风格重写这段话。"

Tips & Limitations

To get the most out of Semantic Router, utilize the trigger_groups_all feature for more flexible, natural language detection. By defining semantic groups, you avoid the need for rigid, exact keyword matching. Note that the router relies on accurate configuration of your model pools; ensure your tasks.json is audited regularly as you add new models to your workspace. The skill performs best when session isolation is strictly enforced via the cron update command for any persistent background tasks.