load available checkpoint 载入可用的模型

This is a checkpoint loader with a filter, and the filter is the whole point. It looks like core ComfyUI's "Load Checkpoint" - pick a model, get MODEL, CLIP, and VAE out - but the dropdown only lists checkpoints that are on the 2lab platform's supported list. The pack ships that list as JSON (standardized/checkpoints.json), and this node shows you only what's on it.

Why would anyone want a loader that hides most of your models? Because of what this pack actually is. comfyUI-tool-2lab lets you publish a workflow to the 2lab cloud so other people can run it as a WeChat mini-program app. When that published workflow runs, it executes on 2lab's servers, which only have the models on their list. If your workflow referenced some obscure checkpoint they don't host, the published app would break. So the "available" loaders force you to build with models the platform can actually run - a guardrail, not a limitation.

The mechanism is straightforward: the node scans your local models/checkpoints/ folder, keeps only filenames that match the standardized list, and presents those as the dropdown. The samples in the schema give you a feel for the list: SDXL stuff like DreamShaper XL (turbo and lightning), Animagine XL, Counterfeit XL, plus SD 1.5 favorites like DreamShaper 8 and the MajicMix series. If you've downloaded the matching files from the pack's Baidu/HuggingFace links, they load like any checkpoint.

The inputs and outputs that matter:

• ckpt_name - the only input, a dropdown of supported checkpoints. Pick from what's there; that's the entire interaction.
• MODEL, CLIP, VAE out - exactly what the core loader gives you, wired into the sampler, the CLIP text encoders, and the VAE decode respectively.

The honest take: if you're not publishing to the 2lab platform, this node is a downgrade from core "Load Checkpoint" - it deliberately restricts your choices, and its selection is a small snapshot of the model world. Use core loader for your own work, and reach for this only when you're building a workflow meant for 2lab publish, where the constraint is doing you a favor. (And note the deeper limitation: the platform only supports this list, so if your dream model isn't on it, that's a conversation with the 2lab team, not a setting.)

Installing it

Part of AI2lab/comfyUI-tool-2lab. ComfyUI Manager → search "comfyUI-tool-2lab" → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/AI2lab/comfyUI-tool-2lab

Restart ComfyUI. No extra dependencies; you may need to download the actual checkpoint files (the standardized list links them), but the node itself doesn't fetch anything.

Common issues

The notable failure mode is a hard error when a checkpoint isn't on the supported list - the node raises something like "checkpoint not in available list, please check model.json." That's by design: it's refusing to let you build a workflow the platform can't run. If the dropdown is empty, your checkpoints folder doesn't have any supported files, or the pack's standardized/checkpoints.json didn't load. And remember the list is the server's catalog, not yours - a model you love may simply never appear here, in which case this node isn't the right tool for your local work.