中文版 | English
AI Agent Skill for Robot Fleet Operations via Feishu/Lark CLI
让 AI Agent 通过飞书 CLI 自动化管理机器人车队：
设备监控 → 多维表格 → 异常告警 → 运维日报
A Feishu CLI Skill that enables AI Agents (Cursor, Claude Code, etc.) to manage robot fleets through Feishu/Lark — no server, no daemon, just a Skill file + CLI commands.
"帮我检查机器人车队状态" → AI reads sensor data → Writes to Bitable → Sends alerts → Creates report
1. Automated Bitable Dashboard
2. Real-time Anomaly Detection & Feishu Interactive Card Alerts
3. AI Predictive Maintenance & Auto Task Dispatch based on history
4. AI-Generated Daily Operations Report
| Feature | Description | Feishu Service |
|---|---|---|
| Status Dashboard | Write device telemetry to Bitable | lark-cli base |
| Anomaly Alerts | Send real-time alerts to group chat | lark-cli im |
| Daily Reports | Generate structured ops reports | lark-cli docs |
| Maintenance Tasks | Create tasks for faulty devices | lark-cli task |
| Scheduled Maintenance | Add calendar events for repairs | lark-cli calendar |
| Fleet Simulator | Zero-dep Python simulator for demos | fleet_simulator.py |
| Type | Condition | Severity |
|---|---|---|
| Overheat | Temperature > 75°C | Critical |
| Low Battery | Battery < 15% | Warning |
| Offline | No heartbeat | Critical |
| CPU Overload | CPU > 90% | Warning |
| Memory Full | Memory > 95% | Warning |
| Weak Signal | WiFi RSSI < -80 dBm | Warning |
| Collision | Emergency stop triggered | Critical |
# Install Feishu CLI (if not already)
npm install -g @larksuite/cli
npx skills add larksuite/cli -y -g
# Install this Skill
npx skills add <your-username>/lark-robot-fleet-ops -y -glark-cli config init --new # Create Feishu app (browser)
lark-cli auth login --domain base,im,docs,task # Authorize permissionspython3 scripts/fleet_simulator.py --count 5 --anomaly-rate 0.3 -o /tmp/fleet_status.jsonOpen Cursor, Claude Code, or any AI tool and say:
请帮我检查机器人车队状态，数据文件在 /tmp/fleet_status.json。 把状态写入飞书多维表格，如果有异常设备发告警到运维群，最后生成运维日报。
The Agent will automatically:
- Read and parse the fleet JSON data
- Write each device's metrics to Bitable
- Detect anomalies and send alerts to your group chat
- Create a structured daily ops report in Feishu Docs
┌─────────────────────────────────────────────────────────┐
│ Data Sources │
│ Simulator │ SSH (Real Devices) │ JSON │
└───────┬─────────────────┬──────────────────────┬────────┘
│ │ │
▼ ▼ ▼
┌─────────────────────────────────────────────────────────┐
│ fleet_collector.sh │
│ (Auto-detect real devices or simulate) │
└───────────────────────┬─────────────────────────────────┘
│ fleet_status.json
▼
┌─────────────────────────────────────────────────────────┐
│ AI Agent + SKILL.md │
│ (Parse data, detect anomalies, orchestrate) │
└──────┬──────────┬──────────┬──────────┬──────────┬──────┘
│ │ │ │ │
▼ ▼ ▼ ▼ ▼
Base IM Docs Task Calendar
(Table) (Alert) (Report) (TODO) (Schedule)
See docs/architecture.md for detailed Mermaid diagrams.
lark-robot-fleet-ops/
├── skills/
│ └── lark-workflow-robot-fleet-ops/
│ └── SKILL.md ← Core: AI Agent workflow instructions
├── scripts/
│ ├── fleet_simulator.py ← Robot fleet data simulator (zero-dep)
│ ├── fleet_collector.sh ← Auto-detect real/sim devices
│ └── setup_base_table.sh ← One-click Bitable setup
├── examples/
│ ├── sample_fleet_status.json ← Demo data (5 devices, 2 anomalies)
│ └── demo_workflow.md ← Step-by-step tutorial
├── docs/
│ └── architecture.md ← Architecture & diagrams
├── package.json ← npm/skills metadata
├── README.md ← You are here
├── README.zh.md ← 中文文档
└── LICENSE ← MIT
The fleet simulator generates realistic telemetry data — no hardware needed:
# Basic: 5 devices, 10% anomaly rate
python3 scripts/fleet_simulator.py
# Custom: 10 devices, 30% anomaly, reproducible
python3 scripts/fleet_simulator.py --count 10 --anomaly-rate 0.3 --seed 42
# Output to file
python3 scripts/fleet_simulator.py -o /tmp/fleet_status.jsonSimulated robot types:
- Companion Robot (
nunu-xxx) — elderly care, home assistant - Patrol Robot (
scout-xxx) — security, surveillance - Delivery Robot (
cargo-xxx) — logistics, warehouse
For production use, create a hosts.txt with your robots' SSH addresses:
# hosts.txt
pi@nunu-001.local
root@192.168.1.50
jetson@orin-nx.local
bash scripts/fleet_collector.sh --hosts hosts.txtThe collector reads system metrics via SSH (/proc, /sys/class/thermal, etc.) and outputs the same JSON format as the simulator.
The project includes advanced data collection and analysis tools:
- MQTT Collector:
python3 scripts/mqtt_collector.py --host broker.hivemq.com - ROS2 (rclpy) Native Node:
python3 scripts/ros2_collector.py(Subscribes to/diagnostics, requires ROS2 environment) - Prometheus Exporter:
python3 scripts/prometheus_exporter.py --port 8000(exposes metrics for Grafana) - Interactive Alert Cards:
examples/alert_card_template.jsonenables rich interactive Feishu alerts - History Trend & AI Predictive Maintenance: AI Agent workflow integrated with Bitable time-series data queries to predict anomalies before they happen.
Thanks to all contributors! The advanced features listed above are now fully implemented. PRs are welcome for these new ideas:
- Enterprise Feishu workflow integrations (e.g. Approvals)
- 3D indoor map integration into Feishu documents
- Multi-lingual and cross-region fleet orchestration
MIT — Built with ❤️ for the Feishu CLI Creator Competition 2026.
- larksuite/cli — The amazing Feishu CLI
- Feishu CLI Creator Competition — Inspiration for this project