basic_Ksampler_simple

If you've ever looked at a "finished" ComfyUI workflow and seen six wires converging on a KSampler node - model, positive, negative, latent, plus whatever else got bolted on - you already get why this node exists. basic_Ksampler_simple is Apt_Preset's answer: the same stock sampler, same math, same output, but everything that used to be four or five separate wires is now one wire called context.

What it actually is

Apt_Preset is built around a "Loader → Controller → Sampler" pipeline, and the glue holding that pipeline together is a custom type called RUN_CONTEXT - a bundle that carries model, conditioning, latent, VAE and clip through the graph as a single connection. Bundling a workflow's plumbing into one wire isn't a new idea in this ecosystem (rgthree's Context nodes do the same job for a different toolkit), but Apt_Preset builds its entire sampler family around it. basic_Ksampler_simple is the plainest member of that family: per the pack's own README, it's meant to be "functionality completely consistent with the official version" - it's not doing anything clever to your image, it's just tidying the wiring.

Inputs and outputs that matter

You don't set much here, which is the point:

• context (required) - the bundle coming out of your loader or the previous controller node. This is where the model, conditioning and latent actually live.
• seed - same as stock KSampler. Fixed seed for reproducible results, random for exploring.
• denoise (0–1, default 1) - how much of the input latent gets overwritten. Leave it at 1 for a normal txt2img pass; drop it for img2img-style partial regeneration.
• image_output - an enum (Hide, Preview, Save, Hide/Save) that folds the usual VAE Decode → Preview/Save Image chain into the node itself. This is the bit stock ComfyUI doesn't give you for free, and it's the main reason to reach for this over a bare KSampler.
• Optional image input, in case you want to feed it a reference image alongside the context.

Outputs are context (pass the bundle on to the next controller/sampler) and image (already decoded, ready to preview or feed downstream).

Installing it

Through ComfyUI Manager: search ComfyUI-Apt_Preset, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/cardenluo/ComfyUI-Apt_Preset.git

then run install.bat (Windows) to pull in the pack's Python dependencies, and restart ComfyUI. This particular node needs nothing extra beyond that - it's a wrapper around ComfyUI's own sampler, not a new model or a heavy dependency.

Common issues

This node is about as low-risk as custom nodes get, since it doesn't touch any external model files. The trouble people run into with Apt_Preset generally happens at the pack level, not this specific node:

Import failures on load. There's at least one report in the wild of ComfyUI-Apt_Preset failing to import on a Windows portable build even after installing through the Manager and via git - the console just shows IMPORT FAILED with no further detail in the log snippet. If that happens to you, check the actual ComfyUI console output (not just the red node) for the missing Python package, run install.bat again, or use Manager's "try fix" on the pack before assuming your workflow itself is broken.

Nothing happens when you hit image_output: Hide. That's intentional - Hide suppresses the preview entirely, which is handy once your graph is dialed in and you don't need to see every intermediate pass. Switch to Preview if you're debugging and want to actually see what came out.

You wired a raw MODEL/LATENT/CONDITIONING into context and it rejected it. context is its own type - you need a loader or controller node upstream (Apt_Preset's own loaders, or chx_input_data if you're bridging from vanilla ComfyUI nodes) that actually outputs RUN_CONTEXT. Standard ComfyUI sampler inputs won't plug straight in here, which is the trade-off for the tidier wiring.