Mingjing — AI Agent Health Center#

Paste this into ClawHub publish page. Replace screenshot URLs with actual images.

---

One-liner#

Zero-dependency, offline-first LLM Agent health monitoring. pip install, zero LLM calls, ~40MB RSS.

Screenshots#

• Health dashboard: [screenshot-url]
• Diagnosis report: [screenshot-url]

Install#

Step 1 — Install backend:

bash

pip install mingjing

Step 2 — Enable probe in OpenClaw:

bash

openclaw plugins install mingjing-probe
openclaw config set plugins.entries.mingjing-probe.enabled true
openclaw gateway

Step 3 — Start web panel:

bash

python3 -m src.ming start --daemon
python3 -m src.ming web start --port 18088

Visit http://localhost:18088 to see live data.

Features#

FeatureDescriptionZero dependenciesPure Python stdlib, no third-party packagesOffline-firstNo outbound traffic, data stored locallyZero LLM cost157 rules, pure rule engine, no model calls7 framework adaptersLangChain / LlamaIndex / CrewAI / OpenHands / Semantic Kernel / AgentScope / Hermes4-level health gradingHealthy / Sub-healthy / Attention / Critical, per instanceWeb dashboardLive event stream, triage coverage, i18n (EN/中文), dark theme

Diagnostic capabilities (157 rules)#

LayerCountCoversSystem20+Memory, IO, CPU, disk, file descriptorsProbe8+Self-check, emit rate, buffer healthAgent20+Orchestration, steps, decisions, rolesModel15+Tokens, latency, output, frequencyTool15+Duration, errors, output, executionNetwork6+Connection, DNS, status codesSecurity5+Injection, sensitive data leaksMemory5+Retrieval, storage, windowingPlugin10+Lifecycle, loading, errors

More#

• Full docs: Mingjing on GitHub
• PyPI: pip install mingjing
• Report issues: https://github.com/wulun811/Ming_qiankun/issues