GPU Pod Management & vLLM Model Deployment

Deploy and manage open-weight LLMs on remote GPU pods via SSH with automatic vLLM configuration.

Mental Model

Local CLI (pi) ──SSH──▶ Remote GPU Pod (Ubuntu + NVIDIA)
                              │
                              ├── vLLM serve (per model)
                              │     └── OpenAI-compatible API (:8001, :8002, ...)
                              │
                              └── ~/.cache/huggingface (model weights)
                                    └── symlinked to persistent storage