Flow-StepPointer

Flow-StepPointer is the dumb half of a two-node system, and it's supposed to be. On its own it does absolutely nothing - it's an output node whose run() returns None, and its only real job is to sit inside a group and act as a landmark. The other half, Flow-Manager, scans the graph for these pointers, reads their numbers and names, and uses them to turn your groups into stages you can run or skip with a button.

If you've used rgthree's Fast Groups Bypasser, you know the general shape of this trick: organize a workflow into groups, then flip whole groups on and off. Flow-StepPointer is the lighter-weight version of the same idea - instead of a dashboard, you mark each stage with a pointer and let Flow-Manager do the rest.

How it works

You place one Flow-StepPointer inside each group you want to treat as a step. When you hit Refresh Steps List on Flow-Manager, it finds every Flow-StepPointer in the graph, reads its two widgets, and figures out which group the pointer is sitting in by testing its position against each group's rectangle. It then sorts the pointers by step number, and those sorted entries become the dropdown Flow-Manager shows you. Position is everything: if a pointer isn't inside any group, Flow-Manager skips it with a warning toast and it never appears in the list.

The two inputs that matter

That's the entire input surface:

• Step (INT, default 0, min 0) - the ordering number. Flow-Manager sorts by this and uses it to decide what "until this step" means, so keep these unique and in the order you want stages to run.
• Name (STRING, default "Step X") - the label shown in Flow-Manager's dropdown, like 1 - Upscale. Set something you'll recognize at a glance.

The output is NULL; the node just needs to exist in the graph. Its required inputs persist in the workflow JSON, so your step numbering and names survive save/reload - that's the one part of this system that isn't ephemeral.

Where people get burned

• Outside a group = invisible. Drop the pointer on the open canvas and Flow-Manager won't even list it. It has to be inside the group it's marking.
• Nested groups are a trap. The code takes the last group that contains the pointer, so nesting gives unpredictable results. Keep groups flat.
• It changes nothing by itself. If you add pointers without a Flow-Manager node in the graph, they're just inert blue squares. They're scaffolding, not a feature.

Install

Same pack, same story as Flow-Manager - no dependencies, no model downloads, no requirements.txt:

cd ComfyUI/custom_nodes
git clone https://github.com/aligjahed/comfy-flow

or search "comfy-flow" in ComfyUI Manager. Restart ComfyUI and the node appears under the Comfy-Flow category. It's a niche pack - barely any community footprint - so if the group-staging workflow isn't your thing, you're not missing a must-have. But if you're debugging multi-stage pipelines, this pair of nodes is a neat, zero-cost way to stop hand-clicking bypass toggles.