AI 異常偵測與人工分類

AI → Human

系統監控指標，AI 分類嚴重程度並路由至人工處理。

5 個節點 · 4 條連接security

eventagenthuman

視覺化

持續監控event

從生產系統串流指標資料。

↓sequential→ 異常偵測

異常偵測agent

ML 模型識別與基準線的偏差。

↓conditional→ 分類嚴重程度

分類嚴重程度system

P1 緊急、P2 高、P3 中、P4 低。

↓sequential→ 發送告警

發送告警api

P1-P2 傳呼值班人員，P3-P4 發送 Slack 通知。

↓sequential→ 工程師決策

工程師決策human

調查、確認或關閉告警。

在編輯器開啟

uc-anomaly-detection.osop.yaml

osop_version: "1.0"
id: "anomaly-detection"
name:"AI 異常偵測與人工分類"
description:"系統監控指標，AI 分類嚴重程度並路由至人工處理。"

nodes:
  - id: "monitor"
    type: "event"
    name: "持續監控"
    description: "從生產系統串流指標資料。"

  - id: "detect"
    type: "agent"
    subtype: "llm"
    name: "異常偵測"
    description: "ML 模型識別與基準線的偏差。"

  - id: "classify"
    type: "system"
    name: "分類嚴重程度"
    description: "P1 緊急、P2 高、P3 中、P4 低。"

  - id: "notify"
    type: "api"
    name: "發送告警"
    description: "P1-P2 傳呼值班人員，P3-P4 發送 Slack 通知。"

  - id: "human_decision"
    type: "human"
    subtype: "review"
    name: "工程師決策"
    description: "調查、確認或關閉告警。"
    security:
      approval_gate: true

edges:
  - from: "monitor"
    to: "detect"
    mode: "sequential"
  - from: "detect"
    to: "classify"
    mode: "conditional"
    when: "anomaly.detected == true"
  - from: "classify"
    to: "notify"
    mode: "sequential"
  - from: "notify"
    to: "human_decision"
    mode: "sequential"