🔗 Egregora Runtime Adapter Merge

Egregora Runtime Adapter Merge is the most honest kind of utility node: it exists so that the pack's other two payload nodes don't force you into a one-thing-only workflow. Where Egregora Spatial Tensor Pack carries images and latents and Egregora Static Payload Pack carries string flags, this node just combines up to three of those adapters into one before you hand it to Egregora Adaptive Diffusion Apply.

Why merging matters

The runtime payload system on Adaptive Diffusion Apply accepts exactly one adapter - there's a single runtime_payload_adapter input, and the mixer validates payload keys for collisions (two payloads trying to claim the same key raises an error rather than silently overwriting). So if you want both spatial guidance and a static flag in the same run, you can't just plug both packs in separately. You merge them first.

That's the entire job: three optional inputs (adapter_a, adapter_b, adapter_c), one runtime_payload_adapter output. Leave any of them unconnected and it merges whatever's there - you don't need all three slots filled. Internally it just combines the underlying payload dictionaries, and the merged adapter still exposes the same build_payload(latent, timestep, cond_dict) contract the mixer expects.

The honest take

If you're not already using the Spatial Tensor Pack or Static Payload Pack, this node is a solution to a problem you don't have. It's the "advanced workflows" branch of the pack - the README calls these nodes out as useful for passing extra runtime tensors, and they exist to keep the denoising wrapper generic instead of hard-coding one workflow's assumptions. For a stock ARMD upscaling or from-scratch generation workflow, you'll never touch it.

When you do need it, it's painless: pack your spatial tensors, pack your flags, merge, feed the result into Adaptive Diffusion Apply. And remember the caveat that applies to the whole payload system - the pack can regionalize compatible spatial tensors and precomputed residual dictionaries, but it doesn't guarantee every native ControlNet chain will work through it. Some Z-Image-specific ControlNet Union setups need separate integration.

Installing

Same as everything else in this pack:

cd ComfyUI/custom_nodes
git clone https://github.com/lucasgattas/ComfyUI-Egregora-ARMD

restart, or install "ComfyUI-Egregora-ARMD" via ComfyUI Manager. No models, no downloads; dependencies are numpy, opencv-python, PyWavelets and torch. From Lucas Gattas (mrgattax), like the rest of the Egregora family - a small, research-flavored, honestly-scoped collection. This node is the definition of "you'll know when you need it."