Dynamic-LoRA-Inference-Server

              Client Request
                    │
                    ▼
             REST / OpenAI API
                    │
            +-------+--------+
            |  Request Router |
            +-------+--------+
                    │
      Adapter ID: "finance-lora"
                    │
                    ▼
         Adapter Cache Manager
      (GPU Cache + CPU Cache + Disk)
                    │
     Loads adapter if not already cached
                    │
                    ▼
    Base LLM (loaded once on GPU)
             +
    Selected LoRA Adapter
                    │
                    ▼
           Generate Response