Guider Switch

Guiders are the modern packaging of "how hard should the sampler follow the prompt." In the advanced sampling world (the "sampler custom" nodes, ComfyUI-Advanced's CFGGuider, or the Flux guidance nodes), a guider is a bundle that wraps a model and a guidance strategy into a single object. The Guider Switch is a two-position toggle that routes one of two GUIDER objects to whichever node consumes it - Input 1 or 2, input1 or input2, that's the whole mechanism.

Why a guider toggle is worth having

If you're switching between classical CFG guidance and a distilled/flow-matching model's "guidance off" mode, you're switching between genuinely different guider bundles, not just different numbers. The whole point of a guider is to encapsulate that difference - and when the difference is encapsulated, it's something you want to swap whole rather than disassemble. Build a CFGGuider on one branch and a guidance-disabled guider on the other, wire both into the switch, and flip a widget instead of rebuilding the sampler feed.

The inputs are forceInput, so both guider objects have to come from upstream nodes that produce GUIDER output - the advanced sampler settings nodes, guider passers, that family. And per the pack's README convention, if the switch itself is bypassed, it passes whatever's on input1 - keep the guider you'd want as the default on slot 1.

A couple of honest notes. First, this node only earns its place if you're already building advanced-sampler workflows; if your graph uses a plain KSampler with a CFG widget, there's no GUIDER object to switch, and this isn't the node you need. Second, like the pack's other 2-way switches, it has a VALIDATE_INPUTS override that always returns True - meaning ComfyUI won't nag about mismatched inputs, so it's on you to wire the right types. The behavior is: empty slots pass None through. That can be a deliberate "disable the guider" switch, or a silent bug source - your call.

What's on the node

• Input - the 1-or-2 selector.
• input1 / input2 - optional GUIDER inputs, wire-only.
• guider output - the chosen GUIDER, feeding your custom sampler node.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2
# restart ComfyUI

Or via ComfyUI Manager, searching "ComfyUI-RvTools_v2". No models to download; the pack's deps are the standard torch/numpy/Pillow/opencv-python/pilgram.

The pack caveat

The README flags RvTools v2 as unmaintained and points to ComfyUI_Eclipse as the successor. The RvTools name also carries scar tissue: the original repo was deleted from GitHub in early 2025, breaking workflows that depended on it and prompting the predictable "can anyone share RvTools?" thread on r/comfyui. v2 is the renamed, still-loading version. For new advanced-sampler builds, Eclipse is where the author moved these nodes; for a legacy workflow that asks for this exact switch, v2 installs and works.