[Internal] Inspect Passthru

[Internal] Inspect Passthru is the smallest node in ComfyUI Functional, and that's about all there is to say at the surface. It takes two inputs and hands them straight back: signal in, signal out; value in, value out. It's a pure relay, generated on the fly by [Internal] Inspect Impl at execution time so that the leaf Output node absorbed by an Inspect can still get its signal and value without the graph getting tangled.

You will never place this node. It exists purely as plumbing. When Inspect Impl compiles your Inspect node, it creates a passthru instance, rewires the absorbed output node's inputs to point at the passthru's sockets, and expands the whole thing back into the graph. The passthru's job is to carry two things in parallel - the ordering signal and the value being probed - into that expanded subgraph without either one tripping over the other. It's the mechanical reason the Inspect docs can promise "lazy, non-eager" debugging: the probed value isn't consumed by anything until this relay hands it to the preview.

It's also a neat demonstration of a ComfyUI pattern this pack leans on constantly: tiny "identity" nodes that wrap a value in a tuple (so ComfyUI treats it as a literal rather than a link) and pass it through. The [Internal] Recover List node does the same trick for list values. In passthru's case the run method literally returns (signal[0], value[0]) - stripping the tuple wrapper off both inputs and forwarding them.

So, the practical takeaway: if you grep a saved workflow and find __InspectPassthru__ sitting there, that's not a node you installed or a dependency you're missing - it's the pack's own scaffolding, and it's a sign the Inspect machinery did its job. Treat it like you'd treat a compiled artifact: don't edit it, don't rearrange it, and if something's wrong, look at the Inspect node upstream rather than the relay.

Install

It comes free with the pack:

cd ComfyUI/custom_nodes && git clone https://github.com/Duanyll/comfyui_functional

then restart ComfyUI - or install Duanyll/comfyui_functional via ComfyUI Manager. No models, no pip requirements, no Basic Data Handling dependency for this one (though you'll want that pack for the functional nodes in general). Since it's generated at runtime, you may never actually see it in the node menu - and that's the point.