Cloud Status

This is the cheapest node in the Cloud Offload pack, and the one you should run first. Cloud Status takes zero inputs, asks the Cloud Offload coordinator what's happening, and returns a JSON string with queue counts, active workers, and your RunPod or Vast.ai balances. That's it. Think of it as a canary you can fire before you queue anything that spends real money.

Why you'd reach for it: because the rest of this pack rents GPUs by the hour, and the worst possible time to discover the coordinator is down is after you've clicked confirm on a rental. Run Cloud Status first. If it comes back with real numbers, your plumbing is alive. If it errors out or returns an empty blob, you caught the problem for free instead of after a paid job submission failed.

How it works

The node is a thin HTTP client. It calls the standalone Cloud Offload coordinator - the separate cloud-offload service that owns all provisioning, queueing, and provider credentials - and prints what it gets back. The pack never imports the coordinator as a library and never touches your provider API keys; credentials live in the coordinator and stay there.

Finding the coordinator follows a fixed order:

1. The CLOUD_OFFLOAD_URL environment variable
2. ~/.cloud-offload/service.json (a JSON file with a url and optional token_path)
3. The localhost default http://127.0.0.1:11435

Port 11434 - Ollama's reserved port - is deliberately never used, so you don't get a silent collision if you run both.

Inputs and outputs

There's exactly one output: status_json (STRING), pretty-printed and key-sorted. Wire it into any text-display node or just read the returned value in the UI. There are no inputs and no provider dropdown - the provider connection is whatever you configured on the coordinator (RunPod is the default; Vast.ai is the alternative).

Install

The pack installs normally, but it's only half the story. The other half is the coordinator, which runs as its own service:

python -m cloud_offload serve --host 127.0.0.1

That command comes from the separate cloud-offload project. The pack itself installs via ComfyUI Manager (search "Cloud Offload"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/jethac/ComfyUI-Cloud-Offload.git

then restart ComfyUI. The pack's own dependencies are trivial - requests and safetensors - with torch, numpy, and Pillow coming from your ComfyUI install. No model downloads, no heavy wheels.

Common issues

The classic failure is an empty or erroring status, which means the coordinator isn't reachable. Walk the discovery chain in order - env var, service file, then port 11435. One deliberate design choice to know about: the pack doesn't health-gate every call, so a coordinator that's mid-restart surfaces as a node error rather than being silently papered over. If queue counts look fine but balances show nothing, your provider credentials aren't connected on the coordinator side yet.

Cloud Offload is brand new - essentially zero community footprint as of this writing - so you won't find a wall of tutorials. The good news is the failure surface here is small and diagnosable, which is exactly what you want from the health check of a tool that can bill you.