Nodes/Shima/Shima Placeholder Tuple
ComfyUI Node

Shima Placeholder Tuple

The Shima Placeholder Tuple is a leftover — know what you're getting

By KDB-USJP·Created 6 months ago·Updated 6 months ago· 2
Shima Placeholder Tuple
    • SDXL_TUPLE
    allow_external_linkingfalse

    Let me save you some time: the Shima Placeholder Tuple is a stub, and I'd read the word "placeholder" in its name literally. It types its single output as SDXL_TUPLE but the actual data it hands you is a tuple of strings built by splitting a hardcoded repr of some long-dead debug session. If you wire that into a real SDXL pipeline you'll get garbage or a crash, not a usable tuple.

    The SDXL_TUPLE type comes from the older Efficiency Nodes / "Efficient Loader" family - it's a pipe that carries (model, clip, positive conditioning, negative conditioning) down one wire so you don't drag four cables across the graph. The Shima pack ships its own more modern equivalent, the BNDL bundle system, and the SDXL_TUPLE compatibility layer here is clearly a legacy leftover. This node's only optional input is allow_external_linking (the island broadcast toggle), and its output is named SDXL_TUPLE.

    What it actually does

    The execute method returns tuple(provided_tuple_string.split(", ")) where provided_tuple_string is a giant literal string pasted into the source - a repr of a (ModelPatcher, CLIP, [[tensor…]], [[tensor…]], None, None, None, None) object from a session that probably ran on someone's Windows box years ago. So you get a Python tuple of strings. It has the right type label for ComfyUI's type checker but none of the right content. It cannot be un-tupled into real model, clip, and conditioning objects.

    What to use instead

    If you need a real pass-through for an SDXL tuple, use Shima SDXL Tuple Pass, which returns the exact tuple you fed it. If you want to build an SDXL pipeline from scratch, use the pack's own bundle generators (Model Citizen, Master Prompt, Latent Maker) and a ReBNDLer. The Placeholder Tuple looks like it was meant as a template node - "here's what an SDXL tuple looks like" - and never got finished or deleted.

    Install

    It's part of the one-pack install:

    cd ComfyUI/custom_nodes
    git clone https://github.com/KDB-USJP/shima_wf.git Shima
    pip install -r requirements.txt
    

    Restart and it appears under Shima/Utilities/Passers. ComfyUI Manager users can just search "Shima". The pack's heavier companions - ComfyUI-Impact-Pack and cg-use-everywhere - are auto-cloned by the installer on first boot because the island system depends on Use Everywhere; that's normal.

    Bottom line

    Don't build anything on this node. It exists, it registers, and it's probably harmless if you leave it sitting on a canvas as a documentation artifact - but the honest answer to "what does it do" is "produces placeholder text with an SDXL_TUPLE label." If you need the real thing, reach for SdxlTuplePass or the BNDL packers. It's a good reminder that even polished packs ship dead code, and that a type label in ComfyUI is a promise the node may not keep.

    CategoryShima/Utilities/Passers

    Inputs (1)

    NameTypeDefaultDescription
    allow_external_linkingoptBOOLEANfalseIf ON, this node broadcasts/receives OUTSIDE the Island (ignores group regex)

    Outputs (1)

    NameTypeDescription
    SDXL_TUPLESDXL_TUPLE