@anjieyang/uncommon-route

English | 简体中文

@anjieyang/uncommon-route

OpenClaw plugin for UncommonRoute, the local LLM router that sends easy requests to cheaper models and saves stronger models for harder work.

If you use OpenClaw and want one local endpoint with smart routing behind it, this plugin is the shortest path.

Mental Model

OpenClaw -> UncommonRoute -> your upstream API

This plugin:

• installs the Python uncommon-route package if needed
• starts uncommon-route serve
• registers the local provider with OpenClaw
• exposes the virtual routing modes like uncommon-route/auto
• keeps request traces and support diagnostics local on disk

Install

openclaw plugins install @anjieyang/uncommon-route
openclaw gateway restart

That is enough to install the plugin.

For real responses, you still need to configure an upstream model API.

Configure An Upstream

UncommonRoute does not host models. It routes to an upstream OpenAI-compatible API.

Example plugin config:

plugins:
  entries:
    "@anjieyang/uncommon-route":
      port: 8403
      upstream: "https://api.commonstack.ai/v1"
      spendLimits:
        hourly: 5.00
        daily: 20.00

Common upstream choices:

Provider	URL
Parallax	http://127.0.0.1:3001/v1
Commonstack	https://api.commonstack.ai/v1
OpenAI	https://api.openai.com/v1
Local Ollama / vLLM	http://127.0.0.1:11434/v1

If your upstream needs a key, set UNCOMMON_ROUTE_API_KEY in the environment where OpenClaw runs.

Parallax is best treated as an experimental local upstream for now: its public docs show POST /v1/chat/completions, but UncommonRoute model discovery may be limited because a public /v1/models route was not obvious in the repo.

What You Get

• a local OpenClaw provider backed by http://127.0.0.1:8403/v1
• uncommon-route/auto for balanced smart routing
• hardcoded additional virtual modes: uncommon-route/fast and uncommon-route/best

The router also keeps a fallback chain, records local feedback, and exposes a local dashboard at http://127.0.0.1:8403/dashboard/.

OpenClaw Commands

Command	Description
/route <prompt>	Preview which model the router would pick
/spend status	Show current spending and limits
/spend set hourly 5.00	Set an hourly spend limit
/feedback <signal>	Use ok, weak, strong, status, or rollback to rate the last routing decision or inspect feedback state

Troubleshooting

If the plugin is installed but responses are failing:

• Make sure your upstream URL is configured.
• Make sure UNCOMMON_ROUTE_API_KEY is set if your provider requires one.
• Open http://127.0.0.1:8403/health.
• Open http://127.0.0.1:8403/dashboard/.
• Run uncommon-route support bundle to export recent traces and diagnostics.

Benchmarks

Current repo benchmarks:

• 82% lower cost than always routing to a premium model
• 93.4% task pass rate on CommonRouterBench
• ~20–25ms warm-process routing overhead on CPU
• 78% tier-match accuracy through the production routing path

Links

• GitHub
• PyPI
• Full README

License

MIT