Init Minio Connector 🐑

This is the node that logs you into your MinIO server, and it's the one you'll place first in every MinIO workflow you build with this pack. It doesn't upload, download, or create anything itself - it just builds a connection object and hands it down the wire to the nodes that do the actual work. Think of it as the login screen before the real interface.

What MinIO actually is, quickly. MinIO is self-hosted, S3-compatible object storage - a private "cloud storage" you run on your own hardware. It's a staple of home-lab AI setups, often sitting next to the box that runs ComfyUI so generated images and training data have somewhere centralized to live instead of filling one machine's disk. Notice the endpoint default is 192.168.31.200:9000 - a LAN address. That tells you exactly who this pack is for: people with a ComfyUI box and a storage server on the same network.

The inputs that matter. Only four, all strings except one:

• endpoint - host and port, like 192.168.31.200:9000. No http:// prefix, just host:port.
• access_key / secret_key - your MinIO credentials (created in the MinIO console).
• secure - a boolean, default false. Leave it off for plain HTTP on your LAN. Flip it on only if your MinIO is behind TLS.

The output. It returns a single minio_connector object of custom type MINIO_CONNECTOR. That's not a standard ComfyUI type - it's this pack's own. You plug that wire into MinioUploadFile, MinioDownloadBucket, or MinioCreateBucketIfNotExists, and they all reuse the same authenticated client. One credential node feeding several transfer nodes is the whole design here.

How it works. On execution the node creates a Minio client from the official minio Python SDK with the credentials you gave it, logs Minio client initialized, and returns the client. Worth knowing: the SDK doesn't actually connect when you construct the client - it's lazy. So a wrong password or an unreachable host won't blow up here. It'll blow up later, on your first real upload or download, with an S3Error. That's a common "why did my workflow pass but the next node fail?" moment.

Install. Search "ComfyUI-MinioConnector" in ComfyUI Manager, install, restart. Or the manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/MieMieeeee/ComfyUI-MinioConnector
cd ComfyUI-MinioConnector
pip install -r requirements.txt

Dependencies are the minio client plus tqdm - light stuff, no models to download, no GPU deps.

Gotchas. Your credentials sit in plaintext inside the workflow JSON. That's fine for a personal LAN workflow, but don't share that workflow file with the keys filled in, and don't expose an unauthenticated ComfyUI to the internet - the whole ecosystem has this problem, not just this pack. Also, if the node doesn't appear after installing, restart ComfyUI completely and check the console for an import error; a failed dependency install will silently drop the whole pack.