ComfyUI GCP Storage Decouple
Custom ComfyUI node that separates storage from compute — read inputs from and write outputs to cloud storage directly.
PythonComfyUIGCPCloud Storage
Overview
A custom ComfyUI node that separates storage from compute. Instead of requiring input/output files to live on the ComfyUI server's local filesystem, this node reads workflow inputs directly from cloud storage (GCP/AWS) and writes results back.
Why?
ComfyUI workflows on cloud GPUs are often ephemeral — the instance boots, runs, and terminates. Local storage doesn't persist. This node solves that by:
- Reading inputs — models, LoRAs, reference images — from GCS or S3 buckets
- Writing outputs — generated images, videos, metadata — back to cloud storage
- Running stateless — the ComfyUI instance needs no persistent volume
Technical Details
- Type: ComfyUI custom node (Python)
- Cloud SDKs: google-cloud-storage, boto3
- Features: Configurable bucket paths, auth via service account / IAM role, automatic retry
- Compatibility: GCP (primary), AWS S3 (secondary)
Status
Beta. Core functionality working; adding directory sync and metadata tagging.