Advanced RAG AI System for Portfolio Support

graph TB
    subgraph "Client Layer"
        A[React Web App
Port 3000]
        B[CLI Interface]
        C[Jupyter Notebook]
        D[HTTP/WebSocket Client]
    end

    subgraph "API Gateway Layer"
        E[Flask + SocketIO
Port 5000]
        F[REST Endpoints]
        G[WebSocket Handler]
    end

    subgraph "Processing Layer"
        H[Advanced RAG Engine]
        I[4 Retrieval Strategies]
        J[Cross-Encoder Reranker]
        K[Memory Manager]
    end

    subgraph "AI/ML Layer"
        L[Ollama LLM
llama2]
        M[HuggingFace Embeddings
all-MiniLM-L6-v2]
        N[Vector Store
ChromaDB + FAISS]
    end

    subgraph "Backend Services"
        O[Express API
Port 3456]
        P[MongoDB]
        Q[Swagger Docs]
    end

    A --> E
    B --> H
    C --> H
    D --> E

    E --> F
    E --> G
    F --> H
    G --> H

    H --> I
    H --> J
    H --> K
    H --> L
    H --> M
    H --> N
    H --> O

    O --> P
    O --> Q

    style A fill:#2196f3,color:#fff
    style E fill:#4caf50,color:#fff
    style H fill:#ff9800,color:#fff
    style L fill:#f44336,color:#fff
    style O fill:#00bcd4,color:#fff
                    

graph TB
    A[User Query] --> B{Query Type Detection}
    
    B -->|Greeting| C[Return Preset Response]
    B -->|Complex Query| D[Document Retrieval]
    
    D --> E[Vector Search
Top-K Similarity]
    E --> F[Entity Extraction]
    F --> G{Extract Entities}
    
    G -->|Person| H[Call Team API]
    G -->|Company| I[Call Investments API]
    G -->|Sector| J[Call Sectors API]
    G -->|URL| K[Call Scrape API]
    
    H --> L[Aggregate API Data]
    I --> L
    J --> L
    K --> L
    
    L --> M[Build LLM Prompt]
    E --> M
    N[Conversation History] --> M
    
    M --> O[Ollama LLM
Generate Response]
    O --> P[Update History]
    P --> Q[Return Response]
    
    C --> Q
    
    style A fill:#e3f2fd,color:#000
    style G fill:#fff3e0,color:#000
    style O fill:#e8f5e9,color:#000
    style Q fill:#f3e5f5,color:#000
                    

graph TD
    START[User Query] --> STRATEGY{Select Strategy}
    
    STRATEGY -->|Semantic| S1[Vector Similarity
ChromaDB Search]
    STRATEGY -->|Hybrid| S2[Ensemble
50% Vector + 50% BM25]
    STRATEGY -->|Multi-Query| S3[Generate Variations
Retrieve All]
    STRATEGY -->|Decomposed| S4[Break into Sub-Queries
Retrieve Each]
    
    S1 --> DOCS1[Retrieved Documents]
    S2 --> DOCS2[Retrieved Documents]
    S3 --> DOCS3[Retrieved Documents]
    S4 --> DOCS4[Retrieved Documents]
    
    DOCS1 --> RERANK[Cross-Encoder
Re-Ranking]
    DOCS2 --> RERANK
    DOCS3 --> RERANK
    DOCS4 --> RERANK
    
    RERANK --> TOP[Top-K Documents
Sorted by Relevance]
    TOP --> LLM[Generate Response
with Citations]
    
    style START fill:#e3f2fd,color:#000
    style STRATEGY fill:#fff3e0,color:#000
    style RERANK fill:#fce4ec,color:#000
    style LLM fill:#e8f5e9,color:#000
                    

mindmap
  root((RAG AI System))
    Frontend
      React 18
      TypeScript 5.7
      Material-UI 6
      Socket.IO Client
      Vite
    Backend API
      Flask 3.1
      Flask-SocketIO
      Flask-CORS
      Python 3.10+
    RAG Engine
      LangChain 0.3
      ChromaDB
      Sentence Transformers
      Cross-Encoders
      BM25Okapi
    AI/ML
      Ollama
      llama2 Model
      all-MiniLM-L6-v2
      ms-marco-MiniLM
    Services
      Express.js
      MongoDB
      Swagger
      Node.js
    DevOps
      Docker
      Docker Compose
      Nginx
      Redis