FDE Ontology — 多源本体自动提取与审核

SkillCloud & infra

Your AI can pull IoT device models straight out of spreadsheets, documents, packet captures, and logs, then audit them and generate code from what it finds. The skill is built on dgiot, an open source industrial IoT platform that works with over 300 device communication formats, including Modbus, OPC UA, and MQTT. Setting up the platform takes about six minutes.

Available today. Use it from your connected AI after setup.

After adding the skill, point your AI at a spreadsheet, document, capture file, or log set that describes your devices. It will extract the device model, audit it, and generate code from it.

Then ask your AI: use the FDE Ontology — 多源本体自动提取与审核 skill

What your AI can do with it

  • Extract IoT device models from spreadsheets, documents, packet captures, and logs
  • Audit extracted device models for problems
  • Generate code from the device models it builds
  • Work with devices that use Modbus, OPC UA, or MQTT
  • Draw on a platform that supports 300+ device communication formats

What this skill tells your AI

The instructions your AI receives, as published by dgiot/dgiot in skills/SKILL.md and read by ahel’s review.

多源输入→本体检出→AI审核→场景升级→代码生成 自动化管线

触发条件

  • "提取本体" / "生成本体" / "梳理本体"
  • "自动创建本体" / "审核本体"
  • "从 Excel/PPT/DOCX 提取设备"
  • "协议解析生成本体"
  • 编辑 thing_model.json / io_ontology.json 后自动审核
  • 对接 fde-toolkit / fde-iot / fde-deploy 输入

输入源→提取器

输入格式提取内容输出
寄存器表.xlsxModbus 地址·类型·单位·阈值thing_model.properties[]
设备清单.docx/.ppt设备名·型号·数量·协议Device[]
网络抓包.pcapA11 5a5a 帧·Modbus TCP·OPC DCOMprotocol.json
配置文件.ini/.conf通道配置·DTU参数·数据库连接Channel[]
架构图.ppt/.drawio网络拓扑·服务器·数据流Site/Gateway[]
进程清单.csv/.txt进程名·版本·内存·心跳Process[]
运行日志.log/.zio设备地址·测点名·实时值Thing[]
投标文件.docx设备材料清单·技术参数Product[]

工作流

Phase 1: EXTRACT — 多源提取
  └── scripts/extract_*.py (per source type)

Phase 2: MERGE — 本体合并
  └── scripts/merge_ontology.py
      去重·冲突检测·补全缺失字段

Phase 3: AUDIT — 7项审核
  └── scripts/audit_ontology.py
      1.寄存器地址冲突  2.数据类型不匹配  3.告警阈值不合理
      4.devaddr重复     5.协议端口冲突    6.TDengine表名长度
      7.ACL规则覆盖检查

Phase 4: UPGRADE — 场景智能升级
  └── scripts/scene_upgrade.py
      读TDengine历史 → 优化阈值·关联规则·预测模型建议

Phase 5: DEPLOY — 入库+生成
  └── 本体 → dgaiot Parse (Site/Gateway/Device/Point)
      ↓
      AI code gen → gen_statem + MQTT + TDengine schema

RULE

  1. 本体是唯一真相源 — 所有输入最终合并为单一本体文件
  2. 审核不可跳过 — 7项检查通过才可进入下一步
  3. 人工确认在审核后、部署前 — AI 建议,人决策
  4. 生产环境运行确定性代码 — 生成的 .erl 编译后直接部署,无 AI 参与
  5. 修正本体不修正代码 — 发现错误改本体,重新生成

输出物

output/
  thing_model.json         合并后的物模型 (带审核标记)
  devices.json             设备注册表
  ontology_report.md       审核报告 (问题清单+处理建议)
  upgrade_plan.md          智能升级方案
  generated/               AI 生成的确定性代码
    dgiot_shadow_guard.erl  编译后的 gen_statem guard
    bridge_config.yaml      MQTT桥接配置
    tdengine_schema.sql     TDengine建表语句

协作技能

fde-ontology ← fde-toolkit (PPT→DLAS)
             ← fde-iot      (pcap→protocol)
             ← fde-deploy   (config→channel)
             ← docx-gen     (投标书→设备清单)
             ↓
             → dgaiot       (本体→代码→部署)

Signals

GitHub stars
5k
Forks
966
Last commit
Sep 2026

ahel recommends instead

Advanced
Catalog kind
skill
Gateway key
skills-dgiot
Source
github.com/dgiot/dgiot