JobbAI:MLOps架构图变更

状态
Tags
Tech_Tag
Created
Jun 23, 2025 01:49 PM
 
graph TD %% v1.0 基础统计版本 subgraph "v1.0: 基础统计分析" direction TB subgraph "应用层v1" A1[职位JD输入] --> A2[TF-IDF计算器] A2 --> A3[统计异常检测] A3 --> A4[控制台输出] end subgraph "数据层v1" A5[最新50个JD<br/>内存存储] end subgraph "基础设施层v1" A6[本地Python脚本<br/>手动执行] end A2 -.-> A5 A6 --> A1 A7[❌ 无历史数据] -.-> A5 A8[❌ 无自动化] -.-> A6 A9[❌ 无语义理解] -.-> A3 end %% 样式定义 classDef v1 fill:#ffebee,stroke:#f44336,stroke-width:2px classDef v2 fill:#e3f2fd,stroke:#2196f3,stroke-width:2px classDef v3 fill:#e8f5e8,stroke:#4caf50,stroke-width:2px classDef business fill:#fff3e0,stroke:#ff9800,stroke-width:2px classDef application fill:#f3e5f5,stroke:#9c27b0,stroke-width:2px classDef platform fill:#e1f5fe,stroke:#00bcd4,stroke-width:2px classDef data fill:#fce4ec,stroke:#e91e63,stroke-width:2px classDef cicd fill:#f1f8e9,stroke:#8bc34a,stroke-width:2px classDef service fill:#fff8e1,stroke:#ffc107,stroke-width:2px classDef monitor fill:#f0f4c3,stroke:#cddc39,stroke-width:2px,stroke-dasharray: 5 5 classDef infra fill:#efebe9,stroke:#795548,stroke-width:2px %% v1.0 class A1,A2,A3,A4,A5,A6,A7,A8,A9 v1 %% v2.0 class B1,B2,B3,B4,B5 application class B6,B7,B8 data class B9,B10,B11 cicd class B12,B13 infra %% v3.0 class C1,C2,C3,C4 business class C5,C6,C7,C8 application class C9,C10,C11,C12 platform class C13,C14,C15,C16 data class C17,C18,C19,C20 cicd class C21,C22,C23 service class C24,C25,C26 monitor class C27,C28,C29 infra
graph TD %% v2.0 MLOps自动化版本 subgraph "v2.0: MLOps自动化流水线" direction TB subgraph "应用层v2" B1[JobDataAccessor<br/>统一数据访问] --> B2[翻译服务HF] B2 --> B3[关键词库匹配] B3 --> B4[TF-IDF + 双语料库] B4 --> B5[异常检测结果] end subgraph "数据层v2" B6[(job_listings<br/>7天完整数据)] B7[(job_analytics_archive<br/>6个月精简数据)] B8[本地HF模型缓存] end subgraph "CI/CD层v2" B9[GitHub Actions<br/>定时任务] B10[翻译流水线<br/>daily@2AM] B11[分析流水线<br/>daily@3AM] end subgraph "基础设施层v2" B12[GitHub Runners<br/>Ubuntu] B13[Supabase数据库<br/>PostgreSQL] end B1 --> B6 B4 --> B7 B9 --> B10 B9 --> B11 B10 --> B2 B11 --> B4 B12 --> B9 B13 --> B6 B13 --> B7 B2 --> B8 end %% 样式定义 classDef v1 fill:#ffebee,stroke:#f44336,stroke-width:2px classDef v2 fill:#e3f2fd,stroke:#2196f3,stroke-width:2px classDef v3 fill:#e8f5e8,stroke:#4caf50,stroke-width:2px classDef business fill:#fff3e0,stroke:#ff9800,stroke-width:2px classDef application fill:#f3e5f5,stroke:#9c27b0,stroke-width:2px classDef platform fill:#e1f5fe,stroke:#00bcd4,stroke-width:2px classDef data fill:#fce4ec,stroke:#e91e63,stroke-width:2px classDef cicd fill:#f1f8e9,stroke:#8bc34a,stroke-width:2px classDef service fill:#fff8e1,stroke:#ffc107,stroke-width:2px classDef monitor fill:#f0f4c3,stroke:#cddc39,stroke-width:2px,stroke-dasharray: 5 5 classDef infra fill:#efebe9,stroke:#795548,stroke-width:2px %% v1.0 class A1,A2,A3,A4,A5,A6,A7,A8,A9 v1 %% v2.0 class B1,B2,B3,B4,B5 application class B6,B7,B8 data class B9,B10,B11 cicd class B12,B13 infra %% v3.0 class C1,C2,C3,C4 business class C5,C6,C7,C8 application class C9,C10,C11,C12 platform class C13,C14,C15,C16 data class C17,C18,C19,C20 cicd class C21,C22,C23 service class C24,C25,C26 monitor class C27,C28,C29 infra
graph TD %% v3.0 LLM语义架构 subgraph "v3.0: LLM混合语义架构" direction TB subgraph "业务层v3" C1[用户招聘需求] --> C2[智能异常分析] C2 --> C3[可解释性报告] C3 --> C4[业务决策支持] end subgraph "应用层v3" C5[EnhancedJobAnomalyDetector<br/>主检测引擎] C6[RoleClassifier<br/>智能角色分类] C7[SemanticAnomalyDetector<br/>语义异常检测] C8[LegacyAnomalyDetector<br/>传统TF-IDF检测] end subgraph "平台层v3" C9[PipelineOrchestrator<br/>流程编排] C10[SemanticClassifier<br/>四轮分级分类] C11[BaselineGenerator<br/>语义基线构建] C12[DatabaseManager<br/>分离存储管理] end subgraph "数据层v3" C13[(semantic_baselines<br/>语义数据表)] C14[(baseline_vectors<br/>向量数据表)] C15[SentenceTransformer<br/>embedding模型缓存] C16[Skills Vocabulary<br/>知识库] end subgraph "CI/CD层v3" C17[GitHub Actions<br/>Serverless Batch] C18[Translation Workflow<br/>daily@2AM] C19[Analysis Workflow<br/>daily@3AM] C20[Baseline Generation<br/>on-demand] end subgraph "模型服务层v3" C21[主服务器<br/>在线Matching服务] C22[批处理服务<br/>GitHub Actions] C23[A/B测试框架<br/>计划中] end subgraph "监控层v3 (计划中)" C24[模型性能监控<br/>🚧 开发中] C25[数据漂移检测<br/>🚧 开发中] C26[业务指标跟踪<br/>🚧 开发中] end subgraph "基础设施层v3" C27[GitHub Runners<br/>批处理计算] C28[Supabase<br/>分布式存储] C29[主服务器<br/>在线服务] end %% 连接关系 C1 --> C5 C5 --> C6 C5 --> C7 C5 --> C8 C6 --> C2 C7 --> C2 C8 --> C2 C2 --> C3 C9 --> C10 C9 --> C11 C9 --> C12 C10 --> C13 C11 --> C13 C11 --> C14 C12 --> C13 C12 --> C14 C16 --> C10 C17 --> C18 C17 --> C19 C17 --> C20 C20 --> C9 C21 --> C5 C22 --> C9 C24 -.-> C5 C25 -.-> C13 C26 -.-> C3 C27 --> C22 C28 --> C13 C28 --> C14 C29 --> C21 C15 --> C7 end %% 样式定义 classDef v1 fill:#ffebee,stroke:#f44336,stroke-width:2px classDef v2 fill:#e3f2fd,stroke:#2196f3,stroke-width:2px classDef v3 fill:#e8f5e8,stroke:#4caf50,stroke-width:2px classDef business fill:#fff3e0,stroke:#ff9800,stroke-width:2px classDef application fill:#f3e5f5,stroke:#9c27b0,stroke-width:2px classDef platform fill:#e1f5fe,stroke:#00bcd4,stroke-width:2px classDef data fill:#fce4ec,stroke:#e91e63,stroke-width:2px classDef cicd fill:#f1f8e9,stroke:#8bc34a,stroke-width:2px classDef service fill:#fff8e1,stroke:#ffc107,stroke-width:2px classDef monitor fill:#f0f4c3,stroke:#cddc39,stroke-width:2px,stroke-dasharray: 5 5 classDef infra fill:#efebe9,stroke:#795548,stroke-width:2px %% v1.0 class A1,A2,A3,A4,A5,A6,A7,A8,A9 v1 %% v2.0 class B1,B2,B3,B4,B5 application class B6,B7,B8 data class B9,B10,B11 cicd class B12,B13 infra %% v3.0 class C1,C2,C3,C4 business class C5,C6,C7,C8 application class C9,C10,C11,C12 platform class C13,C14,C15,C16 data class C17,C18,C19,C20 cicd class C21,C22,C23 service class C24,C25,C26 monitor class C27,C28,C29 infra
 
notion image