graph TB
    User[👤 사용자] --> Web[🌐 웹 인터페이스]

    Web --> Auth[🔐 인증 시스템]
    Web --> Vision[👁️ Vision AI]
    Web --> Recommend[🎯 추천 시스템]
    Web --> Community[💬 커뮤니티]

    Auth --> Supabase[(🐘 Supabase PostgreSQL)]
    Community --> Supabase
    Community --> Storage[📁 Supabase Storage]

    Vision --> Rekognition[🔍 AWS Rekognition]
    Vision --> Gemini[🤖 Gemini 2.0]
    Vision --> Llama[🦙 Llama 3.2]

    Recommend --> OpenAI[🧠 OpenAI Embeddings]
    Recommend --> Qdrant[(🔮 Qdrant Vector DB)]
    Recommend --> GeminiRerank[⚡ Gemini Reranking]

    Sync[🔄 데이터 동기화] --> TourAPI[🗺️ 한국관광공사 API]
    Sync --> MySQL[(🐬 Aiven MySQL)]

    Recommend --> MySQL

    style User fill:#e1f5fe
    style Vision fill:#f3e5f5
    style Recommend fill:#e8f5e8
    style Community fill:#fff3e0

erDiagram
    %% Supabase PostgreSQL - 사용자/커뮤니티
    USERS ||--o{ POSTS : writes
    USERS ||--o{ COMMENTS : writes
    USERS ||--o{ POST_LIKES : likes
    POSTS ||--o{ COMMENTS : has
    POSTS ||--o{ POST_IMAGES : contains
    POSTS ||--o{ POST_LIKES : receives

    %% Aiven MySQL - 여행지 정보
    CONTENT ||--o{ CONTENT_IMAGES : has
    CONTENT ||--|| CONTENT_INTRO : has
    CONTENT ||--|| PET_TOUR_INFO : has
    CONTENT ||--o{ CONTENT_INFOS : has
    CONTENT ||--o{ ROOM_INFOS : has

    %% Vector Database
    VECTOR_EMBEDDINGS ||--|| CONTENT : represents

    USERS {
        uuid id PK
        string username
        string display_name
        string role
        string provider
        timestamp created_at
    }

    CONTENT {
        bigint content_id PK
        int content_type_id
        string title
        string addr1
        point location
        timestamp modified_time
    }

    VECTOR_EMBEDDINGS {
        string id PK
        string content_id FK
        vector embedding
        json metadata
    }

flowchart LR
    A[📷 이미지 업로드] --> B[🔍 이미지 검증]
    B --> C{유효한가?}
    C -->|No| D[❌ 오류 반환]
    C -->|Yes| E[🤖 AWS Rekognition]
    E --> F[📊 라벨 분석]
    F --> G[🐕 동물 종 판별]
    G --> H[💬 중간 결과 반환]
    H --> I[🧠 Gemini 2.0 Flash]
    I --> J{성공?}
    J -->|Yes| K[📝 상세 분석 완료]
    J -->|No| L[🦙 Llama 3.2 Fallback]
    L --> M[📝 분석 완료]
    K --> N[💾 캐시 저장]
    M --> N
    N --> O[📋 최종 보고서]
    style A fill:#e1f5fe
    style I fill:#f3e5f5
    style L fill:#fff3e0
    style O fill:#e8f5e8

flowchart LR
    A[👤 사용자 입력] --> B[🔄 프롬프트 빌딩]
    B --> C[🧠 OpenAI Embeddings]
    C --> D[🔍 벡터 검색]
    D --> E[📍 필터링]
    E --> F[🤖 Gemini 리랭킹]
    F --> G[📋 최종 추천]

    subgraph "Vector Search"
        D --> D1[🔮 Qdrant Query]
        D1 --> D2[📊 Similarity Score]
        D2 --> D3[🎯 Top-K Results]
    end

    subgraph "Intelligent Reranking"
        F --> F1[📝 컨텍스트 분석]
        F1 --> F2[🎯 개인화 점수]
        F2 --> F3[📈 최적 순서]
    end

    style C fill:#e1f5fe
    style D fill:#f3e5f5
    style F fill:#fff3e0