@things-factory/mlops 9.1.19

package/README.md

@@ -0,0 +1,341 @@
+# @things-factory/mlops
+
+**ML lifecycle orchestration platform for Things-Factory**
+
+MLOps 모듈은 머신러닝 모델의 전체 라이프사이클을 관리하는 플랫폼입니다. MLflow, Label Studio 등 다양한 ML 도구들을 통합하여 end-to-end ML 워크플로우를 자동화합니다.
+
+## 핵심 기능
+
+### ✅ Model Lifecycle Management
+- **모델 등록 및 버전 관리** - MLflow와 연동하여 모델 추적
+- **Stage 관리** - None → Staging → Production 프로모션
+- **메트릭 추적** - 정확도, 손실, 커스텀 메트릭 저장
+- **메타데이터 관리** - 프레임워크, 학습 파라미터, 태그
+
+### 🔄 ML Pipeline Orchestration
+- **자동 재학습 파이프라인** - Label Studio 어노테이션 → 자동 학습
+- **스케줄 기반 학습** - Cron 표현식으로 주기적 재학습
+- **이벤트 기반 트리거** - 임계값 도달 시 자동 학습
+- **파이프라인 모니터링** - 실행 이력, 성공률 추적
+
+### 🚀 Model Deployment
+- **환경별 배포** - Development, Staging, Production
+- **배포 상태 추적** - Deploying, Active, Failed, Stopped
+- **성능 메트릭** - Latency, throughput, error rate
+- **롤백 지원** - 문제 발생 시 이전 버전으로 복구
+
+### 📊 Monitoring & Observability
+- **대시보드** - 모델, 파이프라인, 배포 현황 한눈에
+- **실행 이력** - 파이프라인 실행 로그 및 결과
+- **알림** - 파이프라인 실패, 배포 이슈 통보
+
+## 아키텍처
+
+```
+┌─────────────────────────────────────────────────────────┐
+│                  @things-factory/mlops                  │
+│                                                          │
+│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐  │
+│  │   Model      │  │  Pipeline    │  │  Deployment  │  │
+│  │  Registry    │  │  Orchestrator│  │   Manager    │  │
+│  └──────┬───────┘  └──────┬───────┘  └──────┬───────┘  │
+│         │                 │                  │           │
+└─────────┼─────────────────┼──────────────────┼───────────┘
+          │                 │                  │
+  ┌───────▼────────┐ ┌──────▼───────┐  ┌──────▼───────┐
+  │integration-    │ │integration-  │  │  Kubernetes  │
+  │   mlflow       │ │label-studio  │  │  / Lambda    │
+  └────────────────┘ └──────────────┘  └──────────────┘
+```
+
+## 빠른 시작
+
+### 1. 설치
+
+```bash
+cd things-factory
+yarn install
+```
+
+### 2. 모델 등록
+
+```graphql
+mutation RegisterModel {
+  registerMLModel(
+    input: {
+      name: "defect-classifier"
+      version: "v1.0.0"
+      description: "WBM defect classification model"
+      framework: PYTORCH
+      metrics: "{\"accuracy\": 0.95, \"f1\": 0.93}"
+      mlflowRunId: "abc123"
+      labelStudioProjectId: 1
+      tags: "[\"production-ready\", \"wbm\"]"
+    }
+  ) {
+    id
+    name
+    version
+    stage
+  }
+}
+```
+
+### 3. 모델 프로모션
+
+```graphql
+mutation PromoteModel {
+  promoteMLModel(
+    input: { modelId: "uuid", targetStage: PRODUCTION, notes: "Passed QA tests" }
+  ) {
+    id
+    name
+    stage
+  }
+}
+```
+
+### 4. 파이프라인 생성 (향후)
+
+```graphql
+mutation CreatePipeline {
+  createMLPipeline(
+    input: {
+      name: "wbm-auto-retrain"
+      labelStudioProjectId: 1
+      annotationThreshold: 100
+      schedule: "0 2 * * *"
+      promotionThreshold: 0.95
+    }
+  ) {
+    id
+    name
+    status
+  }
+}
+```
+
+## GraphQL API
+
+### Query
+
+```graphql
+# 모든 모델 조회
+query {
+  mlModels {
+    id
+    name
+    version
+    stage
+    framework
+    metrics
+    createdAt
+  }
+}
+
+# Stage별 모델 조회
+query {
+  mlModelsByStage(stage: PRODUCTION) {
+    id
+    name
+    version
+  }
+}
+
+# 모델 버전 조회
+query {
+  mlModelVersions(name: "defect-classifier") {
+    id
+    version
+    stage
+    metrics
+  }
+}
+
+# Label Studio 프로젝트 연결 모델
+query {
+  mlModelsByLabelStudioProject(projectId: 1) {
+    id
+    name
+    version
+  }
+}
+```
+
+### Mutation
+
+```graphql
+# 모델 업데이트
+mutation {
+  updateMLModel(
+    id: "uuid"
+    input: { description: "Updated description", stage: STAGING }
+  ) {
+    id
+    description
+    stage
+  }
+}
+
+# 모델 아카이브
+mutation {
+  archiveMLModel(id: "uuid")
+}
+
+# 모델 삭제
+mutation {
+  deleteMLModel(id: "uuid")
+}
+```
+
+## 엔티티
+
+### MLModel
+- **id**: UUID
+- **name**: 모델 이름
+- **version**: 버전 (semantic versioning 권장)
+- **stage**: None, Staging, Production, Archived
+- **framework**: TensorFlow, PyTorch, Scikit-Learn, etc.
+- **metrics**: JSON 형태의 메트릭
+- **artifactUri**: 모델 파일 저장 위치
+- **mlflowRunId**: MLflow 연동 ID
+- **labelStudioProjectId**: Label Studio 프로젝트 ID
+
+### MLPipeline (향후)
+- **name**: 파이프라인 이름
+- **status**: DRAFT, ACTIVE, PAUSED, ARCHIVED
+- **triggerType**: MANUAL, SCHEDULE, WEBHOOK, EVENT
+- **schedule**: Cron 표현식
+- **annotationThreshold**: 재학습 트리거 임계값
+
+### MLPipelineRun (향후)
+- **pipeline**: 부모 파이프라인
+- **status**: PENDING, RUNNING, SUCCEEDED, FAILED, CANCELLED
+- **startedAt**: 시작 시간
+- **duration**: 실행 시간 (초)
+- **results**: 실행 결과 JSON
+
+### MLDeployment (향후)
+- **model**: 배포된 모델
+- **environment**: DEVELOPMENT, STAGING, PRODUCTION
+- **status**: DEPLOYING, ACTIVE, FAILED, STOPPED
+- **endpoint**: API 엔드포인트 URL
+
+## 사용 예제
+
+### WBM 자동 재학습 워크플로우
+
+```typescript
+import { registerMLModel, promoteMLModel } from '@things-factory/mlops'
+import { getAnnotations } from '@things-factory/integration-label-studio'
+
+// 1. Label Studio에서 새 어노테이션 확인
+const annotations = await getAnnotations(projectId)
+
+if (annotations.length > lastCount + 100) {
+  // 2. 재학습 실행 (외부 스크립트)
+  const trainResult = await runTraining(annotations)
+
+  // 3. MLOps에 모델 등록
+  const model = await registerMLModel({
+    name: 'wbm-classifier',
+    version: `v${Date.now()}`,
+    framework: 'PYTORCH',
+    metrics: JSON.stringify(trainResult.metrics),
+    mlflowRunId: trainResult.runId,
+    labelStudioProjectId: projectId
+  })
+
+  // 4. 성능 기준 충족 시 자동 프로모션
+  if (trainResult.metrics.accuracy > 0.95) {
+    await promoteMLModel({
+      modelId: model.id,
+      targetStage: 'PRODUCTION'
+    })
+  }
+}
+```
+
+## 통합 모듈
+
+### @things-factory/integration-mlflow
+- MLflow experiment tracking
+- Model registry 연동
+- Artifact 저장
+
+### @things-factory/integration-label-studio
+- 어노테이션 데이터 소싱
+- 재학습 트리거
+- 데이터 품질 관리
+
+## 로드맵
+
+### Phase 1: 기본 모델 관리 (현재)
+- ✅ 모델 등록/조회
+- ✅ 버전 관리
+- ✅ Stage 프로모션
+- ✅ GraphQL API
+
+### Phase 2: 파이프라인 자동화
+- ⏳ 자동 재학습 파이프라인
+- ⏳ 스케줄링
+- ⏳ 이벤트 트리거
+
+### Phase 3: 배포 관리
+- ⏳ 모델 서빙
+- ⏳ A/B 테스트
+- ⏳ 카나리 배포
+
+### Phase 4: 고급 기능
+- ⏳ 모델 드리프트 감지
+- ⏳ 자동 롤백
+- ⏳ AutoML 통합
+
+## 파일 구조
+
+```
+mlops/
+├── server/
+│   ├── entities/
+│   │   ├── ml-model.ts              # 모델 엔티티
+│   │   ├── ml-pipeline.ts           # 파이프라인 엔티티
+│   │   ├── ml-pipeline-run.ts       # 실행 이력 엔티티
+│   │   └── ml-deployment.ts         # 배포 엔티티
+│   ├── service/
+│   │   ├── ml-model/
+│   │   │   ├── ml-model-query.ts    # 모델 조회
+│   │   │   ├── ml-model-mutation.ts # 모델 변경
+│   │   │   └── ml-model-types.ts    # GraphQL 타입
+│   │   ├── ml-pipeline/
+│   │   ├── ml-pipeline-run/
+│   │   └── ml-deployment/
+│   ├── routes.ts
+│   └── index.ts
+├── client/
+│   ├── pages/
+│   │   ├── ml-model/
+│   │   │   └── ml-model-list-page.ts
+│   │   ├── ml-pipeline/
+│   │   │   └── ml-pipeline-list-page.ts
+│   │   ├── ml-deployment/
+│   │   │   └── ml-deployment-list-page.ts
+│   │   └── mlops-dashboard/
+│   │       └── mlops-dashboard-page.ts
+│   ├── route.ts
+│   ├── bootstrap.ts
+│   └── index.ts
+├── translations/
+│   ├── en.json
+│   └── ko.json
+├── package.json
+├── things-factory.config.js
+└── README.md
+```
+
+## 라이선스
+
+MIT
+
+## 기여
+
+버그 리포트 및 기능 제안은 GitHub Issues로 부탁드립니다.

package/client/bootstrap.ts

@@ -0,0 +1,41 @@
+import { store } from '@operato/shell'
+
+export default function bootstrap() {
+  /**
+   * MLOps Module Bootstrap
+   *
+   * Initialize module-specific settings, event listeners, etc.
+   */
+
+  // Example: Register menu items
+  store.dispatch({
+    type: 'ADD_MENU',
+    menu: {
+      category: 'mlops',
+      name: 'mlops',
+      icon: 'model_training',
+      childrens: [
+        {
+          name: 'ml-models',
+          path: 'ml-model-list',
+          icon: 'psychology'
+        },
+        {
+          name: 'ml-pipelines',
+          path: 'ml-pipeline-list',
+          icon: 'account_tree'
+        },
+        {
+          name: 'ml-deployments',
+          path: 'ml-deployment-list',
+          icon: 'rocket_launch'
+        },
+        {
+          name: 'mlops-dashboard',
+          path: 'mlops-dashboard',
+          icon: 'dashboard'
+        }
+      ]
+    }
+  })
+}

package/client/index.ts

@@ -0,0 +1,2 @@
+export { default as route } from './route'
+export { default as bootstrap } from './bootstrap'

package/client/pages/ml-deployment/ml-deployment-list-page.ts

@@ -0,0 +1,46 @@
+import { css, html, LitElement } from 'lit'
+import { customElement } from 'lit/decorators.js'
+
+@customElement('ml-deployment-list-page')
+export class MLDeploymentListPage extends LitElement {
+  static styles = [
+    css`
+      :host {
+        display: flex;
+        flex-direction: column;
+        height: 100%;
+        padding: 20px;
+      }
+
+      h2 {
+        margin: 0 0 20px 0;
+      }
+
+      .placeholder {
+        text-align: center;
+        padding: 40px;
+        color: #666;
+      }
+    `
+  ]
+
+  render() {
+    return html`
+      <h2>ML Deployments</h2>
+
+      <div class="placeholder">
+        <p>ML Deployment list will be displayed here</p>
+        <p>
+          This page will show:
+          <ul style="text-align: left; display: inline-block;">
+            <li>Active deployments</li>
+            <li>Environment (Development, Staging, Production)</li>
+            <li>Deployment status</li>
+            <li>Performance metrics (latency, error rate)</li>
+            <li>Actions: Stop, Restart, Rollback</li>
+          </ul>
+        </p>
+      </div>
+    `
+  }
+}

package/client/pages/ml-model/ml-model-list-page.ts

@@ -0,0 +1,46 @@
+import { css, html, LitElement } from 'lit'
+import { customElement } from 'lit/decorators.js'
+
+@customElement('ml-model-list-page')
+export class MLModelListPage extends LitElement {
+  static styles = [
+    css`
+      :host {
+        display: flex;
+        flex-direction: column;
+        height: 100%;
+        padding: 20px;
+      }
+
+      h2 {
+        margin: 0 0 20px 0;
+      }
+
+      .placeholder {
+        text-align: center;
+        padding: 40px;
+        color: #666;
+      }
+    `
+  ]
+
+  render() {
+    return html`
+      <h2>ML Models</h2>
+
+      <div class="placeholder">
+        <p>ML Model list will be displayed here</p>
+        <p>
+          This page will show:
+          <ul style="text-align: left; display: inline-block;">
+            <li>Registered ML models</li>
+            <li>Model versions</li>
+            <li>Deployment stages (None, Staging, Production)</li>
+            <li>Model metrics</li>
+            <li>Actions: Promote, Archive, Delete</li>
+          </ul>
+        </p>
+      </div>
+    `
+  }
+}

package/client/pages/ml-pipeline/ml-pipeline-list-page.ts

@@ -0,0 +1,46 @@
+import { css, html, LitElement } from 'lit'
+import { customElement } from 'lit/decorators.js'
+
+@customElement('ml-pipeline-list-page')
+export class MLPipelineListPage extends LitElement {
+  static styles = [
+    css`
+      :host {
+        display: flex;
+        flex-direction: column;
+        height: 100%;
+        padding: 20px;
+      }
+
+      h2 {
+        margin: 0 0 20px 0;
+      }
+
+      .placeholder {
+        text-align: center;
+        padding: 40px;
+        color: #666;
+      }
+    `
+  ]
+
+  render() {
+    return html`
+      <h2>ML Pipelines</h2>
+
+      <div class="placeholder">
+        <p>ML Pipeline list will be displayed here</p>
+        <p>
+          This page will show:
+          <ul style="text-align: left; display: inline-block;">
+            <li>Configured ML pipelines</li>
+            <li>Pipeline status (Active, Paused, Draft)</li>
+            <li>Trigger configuration (Manual, Schedule, Event)</li>
+            <li>Last execution time</li>
+            <li>Actions: Run, Pause, Edit, Delete</li>
+          </ul>
+        </p>
+      </div>
+    `
+  }
+}

package/client/pages/mlops-dashboard/mlops-dashboard-page.ts

@@ -0,0 +1,90 @@
+import { css, html, LitElement } from 'lit'
+import { customElement } from 'lit/decorators.js'
+
+@customElement('mlops-dashboard-page')
+export class MLOpsDashboardPage extends LitElement {
+  static styles = [
+    css`
+      :host {
+        display: flex;
+        flex-direction: column;
+        height: 100%;
+        padding: 20px;
+      }
+
+      h2 {
+        margin: 0 0 20px 0;
+      }
+
+      .dashboard-grid {
+        display: grid;
+        grid-template-columns: repeat(auto-fit, minmax(300px, 1fr));
+        gap: 20px;
+        margin-top: 20px;
+      }
+
+      .dashboard-card {
+        background: white;
+        border: 1px solid #e0e0e0;
+        border-radius: 8px;
+        padding: 20px;
+        box-shadow: 0 2px 4px rgba(0, 0, 0, 0.1);
+      }
+
+      .dashboard-card h3 {
+        margin: 0 0 15px 0;
+        font-size: 18px;
+      }
+
+      .placeholder {
+        text-align: center;
+        padding: 40px;
+        color: #666;
+      }
+    `
+  ]
+
+  render() {
+    return html`
+      <h2>MLOps Dashboard</h2>
+
+      <div class="dashboard-grid">
+        <div class="dashboard-card">
+          <h3>Models Overview</h3>
+          <div class="placeholder">
+            <p>Total models: -</p>
+            <p>Production: -</p>
+            <p>Staging: -</p>
+          </div>
+        </div>
+
+        <div class="dashboard-card">
+          <h3>Pipeline Status</h3>
+          <div class="placeholder">
+            <p>Active pipelines: -</p>
+            <p>Recent runs: -</p>
+            <p>Success rate: -</p>
+          </div>
+        </div>
+
+        <div class="dashboard-card">
+          <h3>Deployment Health</h3>
+          <div class="placeholder">
+            <p>Active deployments: -</p>
+            <p>Avg latency: -</p>
+            <p>Error rate: -</p>
+          </div>
+        </div>
+
+        <div class="dashboard-card">
+          <h3>Recent Activity</h3>
+          <div class="placeholder">
+            <p>Recent model registrations</p>
+            <p>Recent pipeline runs</p>
+            <p>Recent promotions</p>
+          </div>
+        </div>
+      </div>
+    `
+  }
+}

package/client/route.ts

@@ -0,0 +1,19 @@
+export default function route(page: string) {
+  switch (page) {
+    case 'ml-model-list':
+      import('./pages/ml-model/ml-model-list-page')
+      return page
+
+    case 'ml-pipeline-list':
+      import('./pages/ml-pipeline/ml-pipeline-list-page')
+      return page
+
+    case 'ml-deployment-list':
+      import('./pages/ml-deployment/ml-deployment-list-page')
+      return page
+
+    case 'mlops-dashboard':
+      import('./pages/mlops-dashboard/mlops-dashboard-page')
+      return page
+  }
+}

package/client/tsconfig.json

@@ -0,0 +1,12 @@
+{
+  "extends": "../../tsconfig-base.json",
+  "compilerOptions": {
+    "experimentalDecorators": true,
+    "strict": true,
+    "declaration": true,
+    "module": "esnext",
+    "outDir": "../dist-client",
+    "baseUrl": "./"
+  },
+  "include": ["./**/*"]
+}

package/dist-client/bootstrap.d.ts

@@ -0,0 +1 @@
+export default function bootstrap(): void;

package/dist-client/bootstrap.js

@@ -0,0 +1,40 @@
+import { store } from '@operato/shell';
+export default function bootstrap() {
+    /**
+     * MLOps Module Bootstrap
+     *
+     * Initialize module-specific settings, event listeners, etc.
+     */
+    // Example: Register menu items
+    store.dispatch({
+        type: 'ADD_MENU',
+        menu: {
+            category: 'mlops',
+            name: 'mlops',
+            icon: 'model_training',
+            childrens: [
+                {
+                    name: 'ml-models',
+                    path: 'ml-model-list',
+                    icon: 'psychology'
+                },
+                {
+                    name: 'ml-pipelines',
+                    path: 'ml-pipeline-list',
+                    icon: 'account_tree'
+                },
+                {
+                    name: 'ml-deployments',
+                    path: 'ml-deployment-list',
+                    icon: 'rocket_launch'
+                },
+                {
+                    name: 'mlops-dashboard',
+                    path: 'mlops-dashboard',
+                    icon: 'dashboard'
+                }
+            ]
+        }
+    });
+}
+//# sourceMappingURL=bootstrap.js.map

package/dist-client/bootstrap.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"bootstrap.js","sourceRoot":"","sources":["../client/bootstrap.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,KAAK,EAAE,MAAM,gBAAgB,CAAA;AAEtC,MAAM,CAAC,OAAO,UAAU,SAAS;IAC/B;;;;OAIG;IAEH,+BAA+B;IAC/B,KAAK,CAAC,QAAQ,CAAC;QACb,IAAI,EAAE,UAAU;QAChB,IAAI,EAAE;YACJ,QAAQ,EAAE,OAAO;YACjB,IAAI,EAAE,OAAO;YACb,IAAI,EAAE,gBAAgB;YACtB,SAAS,EAAE;gBACT;oBACE,IAAI,EAAE,WAAW;oBACjB,IAAI,EAAE,eAAe;oBACrB,IAAI,EAAE,YAAY;iBACnB;gBACD;oBACE,IAAI,EAAE,cAAc;oBACpB,IAAI,EAAE,kBAAkB;oBACxB,IAAI,EAAE,cAAc;iBACrB;gBACD;oBACE,IAAI,EAAE,gBAAgB;oBACtB,IAAI,EAAE,oBAAoB;oBAC1B,IAAI,EAAE,eAAe;iBACtB;gBACD;oBACE,IAAI,EAAE,iBAAiB;oBACvB,IAAI,EAAE,iBAAiB;oBACvB,IAAI,EAAE,WAAW;iBAClB;aACF;SACF;KACF,CAAC,CAAA;AACJ,CAAC","sourcesContent":["import { store } from '@operato/shell'\n\nexport default function bootstrap() {\n  /**\n   * MLOps Module Bootstrap\n   *\n   * Initialize module-specific settings, event listeners, etc.\n   */\n\n  // Example: Register menu items\n  store.dispatch({\n    type: 'ADD_MENU',\n    menu: {\n      category: 'mlops',\n      name: 'mlops',\n      icon: 'model_training',\n      childrens: [\n        {\n          name: 'ml-models',\n          path: 'ml-model-list',\n          icon: 'psychology'\n        },\n        {\n          name: 'ml-pipelines',\n          path: 'ml-pipeline-list',\n          icon: 'account_tree'\n        },\n        {\n          name: 'ml-deployments',\n          path: 'ml-deployment-list',\n          icon: 'rocket_launch'\n        },\n        {\n          name: 'mlops-dashboard',\n          path: 'mlops-dashboard',\n          icon: 'dashboard'\n        }\n      ]\n    }\n  })\n}\n"]}

package/dist-client/index.d.ts

@@ -0,0 +1,2 @@
+export { default as route } from './route';
+export { default as bootstrap } from './bootstrap';