Checkpoint Loader w/Name (WLSH)

Your standard checkpoint loader hands you MODEL, CLIP, and VAE - everything you need to actually generate an image, but nothing that tells the rest of your graph which checkpoint you loaded. This node is the same loader with one extra output bolted on: the checkpoint's filename, as a plain string, so anything downstream that wants to know or record which model made an image can just ask this node instead of you retyping the name somewhere else.

That sounds minor until you're trying to keep track of outputs across a session where you're swapping checkpoints to compare results. WLSH's own save nodes have a modelname field specifically for embedding which checkpoint produced an image into the saved metadata - but without a node like this one, populating that field means manually typing the checkpoint name as a separate string, which drifts out of sync the moment you switch models in the loader and forget to update the text box to match. Wire this node's modelname output straight into your save node instead, and it's always accurate because it's coming from the same place the model itself is loading from.

Input: ckpt_name, an enum populated from whatever .safetensors or .ckpt files you have in your ComfyUI checkpoints folder - functionally identical to the dropdown on the stock checkpoint loader, just pulling from the same source.

Outputs, four of them: MODEL, CLIP, and VAE behave exactly like the stock loader's outputs - wire them into your sampler, your text encoder, and your VAE decode/encode nodes as normal. modelname is the addition: a STRING output carrying the loaded checkpoint's filename, ready to feed into anything that wants it - most usefully, WLSH's own save nodes' modelname input, or Build Filename String (WLSH) if you want the model name baked into a filename elsewhere.

Installing it: through ComfyUI Manager, search "wlsh_nodes"; or clone the pack manually:

cd ComfyUI/custom_nodes
git clone https://github.com/wallish77/wlsh_nodes

Restart ComfyUI afterward. No model downloads needed for the node itself - same as any checkpoint loader, the actual checkpoint files need to already be in your models/checkpoints folder for the dropdown to show anything.

Troubleshooting: if the ckpt_name dropdown is empty, that's not this node's fault - it means ComfyUI can't find any checkpoints in the expected folder, same failure you'd get with the stock loader, and the fix is making sure your .safetensors files are actually in models/checkpoints (or wherever your extra_model_paths.yaml points ComfyUI to look). If modelname is coming through with an unexpected extension or path prefix attached rather than a clean name, remember it's reflecting whatever the raw filename actually is on disk - renaming the checkpoint file itself is the fix, not something to chase in the node. And if you swap this node in mid-workflow to replace a stock checkpoint loader, double check you've re-wired all three model outputs (MODEL/CLIP/VAE) to their original destinations - it's an easy step to miss since the node looks nearly identical to the one you're replacing.