Deprecated · Sampling Trace Model Snapshot

Sampling Trace Model Snapshot takes a point-in-time picture of your MODEL's state and passes the model through unchanged. The picture captures patch count, active transformer patch types, wrappers, and callbacks - the invisible machinery that LoRAs, IPAdapter, and ControlNet hang off. It's the "what is actually loaded on this model right now" question, answered in the middle of your graph.

This is the probe version of something the pack's main flow already does. When you use the one-node setup or the deprecated Sampling Trace Model, a model snapshot is captured automatically as part of the run so the A/B comparison can show patch differences between two generations. This standalone probe is for the moments you want that snapshot at a specific point - say, verifying that your LoRA actually landed on the model before it reaches the sampler, or checking whether a patch survived a model-merge node. It records and moves on; it does not start sampling trace capture by itself, and the node description says so explicitly.

It's one of the pack's five hidden probes - deprecated and hidden from normal search in 0.4.0, kept for saved-workflow compatibility and internal diagnostics.

Inputs and outputs

• model (required, MODEL) - the branch to snapshot.
• label (optional, STRING) - a name for the probe in the panel.
• trace_session (optional, TRACE_SESSION) - connect one to store the snapshot into a run's evidence; otherwise it's a live event.
• Output: model, unchanged.

Install

git clone https://github.com/ILcrowe/ComfyUI-SamplingTrace-Inspector.git ComfyUI/custom_nodes/ComfyUI-SamplingTrace-Inspector

or ComfyUI Manager → "Sampling Trace", restart. No dependencies or model downloads.

Where people get burned

• Hidden from search. Deprecated by design; it loads from old saved workflows.
• "Does this start my trace?" No - and that's intentional. It's a probe, not the trace entry point. If you want a traced run, use the one-node setup.
• Snapshot ≠ prevention. It records what's on the model; it doesn't fix a missing patch. If your LoRA isn't on the model, this tells you - then you go fix the wiring.

For people debugging patch-heavy workflows on old graphs, this is the closest thing the pack offers to a "did my stack actually apply?" check. That's worth having in your pocket even if you rarely reach for it.