claw-turbo
https://github.com/jacobye2017-afk/claw-turbo
Zero-latency, zero-ML skill routing middleware for OpenClaw.
Sits between OpenClaw and Ollama as an HTTP proxy. Intercepts user messages matching regex patterns and executes skill scripts directly — the LLM never sees them.
Problem it solves: Local LLMs (Gemma 4 26B, etc.) are inconsistent at following SKILL.md instructions for simple commands. They add wrong flags, strip keywords, and waste context window on trivial routing.
Solution: Pattern-matched commands bypass the LLM entirely. <1ms, 100% accurate. Complex queries pass through to the LLM normally.
# routes.yaml - add regex rules for any skill
routes:
- name: print-container
patterns:
- '打印\s*(?P<container>[A-Z]{4}\d{7})\s*(的)?\s*(资料|全部)'
command: 'bash /path/to/run.sh "{{raw_message}}"'
- Zero ML dependencies (PyYAML only)
- Hot-reload routes
- CLI for testing:
claw-turbo test "your message"
- MIT License
claw-turbo
https://github.com/jacobye2017-afk/claw-turbo
Zero-latency, zero-ML skill routing middleware for OpenClaw.
Sits between OpenClaw and Ollama as an HTTP proxy. Intercepts user messages matching regex patterns and executes skill scripts directly — the LLM never sees them.
Problem it solves: Local LLMs (Gemma 4 26B, etc.) are inconsistent at following SKILL.md instructions for simple commands. They add wrong flags, strip keywords, and waste context window on trivial routing.
Solution: Pattern-matched commands bypass the LLM entirely. <1ms, 100% accurate. Complex queries pass through to the LLM normally.
claw-turbo test "your message"