🔢 AGSoft Loop Float

AGSoft Loop Float generates a sequence of floats between a start and end value - with three things people actually care about in a float range: the end value is included, the step can be fractional, and you can control the decimal precision. It's the float counterpart to AGSoft Loop Integer, and it's the kind of node you grab when you want to sweep a parameter smoothly instead of guessing three values by hand.

Float ranges are where naive loop logic falls apart. range() in Python can't do fractional steps, floating-point error turns 0.1 + 0.2 into a coin flip, and most "step through these values" tools quietly stop before the end value. This node solves all three. If you've ever tried to sweep CFG from 4.0 to 8.0 in 0.5 steps and gotten a list that ends at 7.5, you know exactly why this exists.

How it works

• start_at_step - first value (default 0). Can be negative, so descending sweeps are just start > end.
• end_at_step - last value, and it is included. start=1.555, end=2.000, jump=0.222 gives you [1.555, 1.777, 2.000], per the node's own example.
• jump - the step, and it can be fractional (min 0.001). Unlike the integer version, sub-unit steps are the whole point here.
• precision - 1, 2, or 3 decimal places. The node rounds each value with proper half-up rounding rather than letting float error accumulate. Set precision to 1 and start=0.05 becomes 0.1; set 3 and you can walk values like 0.125 cleanly.

Output: a single FLOAT list, ready to feed a loop sampler, a scheduler, or any list-driven node. The rounding is deliberate - an animation or CFG schedule with 0.30000000000000004 in it is a debugging nightmare, and this node just doesn't produce those.

Installation

From comfyui-AGSoft:

cd ComfyUI/custom_nodes
git clone https://github.com/Art-xmaster/comfyui-AGSoft.git
# restart ComfyUI

Or via ComfyUI Manager (search "comfyui-AGSoft"). No models, no heavy deps.

Where it fits

Reach for it when you want a sweep - CFG scales, denoise ramps, LoRA strength walks - and let AGSoft Loop Fix Float handle the case where you want to hand-type specific values instead. The inclusive end is the feature to remember: most "range" utilities in this ecosystem stop one short of your target, and then your loop mysteriously never tests the value you actually wanted to test.