Brunobkr/llama.cpp_AlgMor24_github

1```mermaid2%% ROUTER Mode Data Flow (multi-model)3%% Detailed flows: ./flows/server-flow.mmd, ./flows/models-flow.mmd, ./flows/chat-flow.mmd4 5sequenceDiagram6    participant User as 👤 User7    participant UI as 🧩 UI8    participant Stores as 🗄️ Stores9    participant DB as 💾 IndexedDB10    participant API as 🌐 llama-server11 12    Note over User,API: 🚀 Initialization (see: server-flow.mmd, models-flow.mmd)13 14    UI->>Stores: initialize()15    Stores->>DB: load conversations16    Stores->>API: GET /props17    API-->>Stores: {role: "router"}18    Stores->>API: GET /v1/models19    API-->>Stores: models[] with status (loaded/available)20    loop each loaded model21        Stores->>API: GET /props?model=X22        API-->>Stores: modalities (vision/audio)23    end24 25    Note over User,API: 🔄 Model Selection (see: models-flow.mmd)26 27    User->>UI: select model28    alt model not loaded29        Stores->>API: POST /models/load30        loop poll status31            Stores->>API: GET /v1/models32            API-->>Stores: check if loaded33        end34        Stores->>API: GET /props?model=X35        API-->>Stores: cache modalities36    end37    Stores->>Stores: validate modalities vs conversation38    alt valid39        Stores->>Stores: select model40    else invalid41        Stores->>API: POST /models/unload42        UI->>User: show error toast43    end44 45    Note over User,API: 💬 Chat Flow (see: chat-flow.mmd)46 47    User->>UI: send message48    UI->>Stores: sendMessage()49    Stores->>DB: save user message50    Stores->>API: POST /v1/chat/completions {model: X}51    Note right of API: router forwards to model52    loop streaming53        API-->>Stores: SSE chunks + model info54        Stores-->>UI: reactive update55    end56    API-->>Stores: done + timings57    Stores->>DB: save assistant message + model used58 59    Note over User,API: 🔁 Regenerate (optional: different model)60 61    User->>UI: regenerate62    Stores->>Stores: validate modalities up to this message63    Stores->>DB: create message branch64    Note right of Stores: same streaming flow65 66    Note over User,API: ⏹️ Stop67 68    User->>UI: stop69    Stores->>Stores: abort stream70    Stores->>DB: save partial response71 72    Note over User,API: 🗑️ LRU Unloading73 74    Note right of API: Server auto-unloads LRU models<br/>when cache full75    User->>UI: select unloaded model76    Note right of Stores: triggers load flow again77```78