greencompute110/greencompute-node ? reverse-engineered prompt

Build me a single miner node service for the Green Compute Bittensor subnet that I can run on a Linux machine with NVIDIA GPUs. I want one process that registers with the validator, reports my GPU capacity, polls for leases, and then automatically handles the different kinds of work a miner can get. That means serving assigned catalog inference models, letting users deploy private Hugging Face models on my GPUs, creating rentable SSH pod containers, and leaving a basic VM path in place even if that part is just a stub for now.

Please make it easy to start with Docker Compose, using a .env file for things like validator URLs, my hotkey and wallet names, GPU model and VRAM, public SSH host, and an HF token. It should expose a simple readiness check on port 8007, keep sending heartbeats, track lease state changes, and proxy inference requests correctly, including the validator nonce behavior so probes pass. If anything is unclear, check current docs online and wire it up in a practical production ready way.