LoRA Compatibility Analyzer

Every other node in this pack either merges LoRAs or configures how the merge happens. This one is different - it doesn't merge anything. It's a planning tool: point it at a stack and it tells you how well those LoRAs are likely to get along, before you commit to running an actual merge.

What it is and why you'd use it

Useful in two moments. First, when you're about to build a merge and want a sanity check before spending compute on a full LoRA Optimizer or AutoTuner run - is this pairing even promising, or are you about to fight a losing battle? Second, when a merge came out badly and you want to understand why without re-running the whole pipeline in debug mode. The node runs the same underlying analysis the Optimizer uses internally (pairwise conflict ratios, cosine similarity, magnitude distribution, key overlap between LoRAs) but stops there instead of proceeding to a merge - you get the diagnosis without the treatment.

How it works

It's toggled off by default (enabled=false), and deliberately so - the README's own framing is that this keeps normal queue execution cheap; you don't want a compatibility analysis silently running every time you execute a workflow that happens to have this node sitting in it. Flip enabled on when you actually want the check to run. There's also a genuinely unusual feature here for a ComfyUI node: create_nodes, on by default, which can automatically build LoRA Stack and Load LoRA nodes on your canvas based on what the analysis suggests - rather than just reporting numbers and leaving you to act on them manually.

The inputs and outputs that matter

• enabled (default false) - the master switch. Nothing runs unless this is on.
• create_nodes (default true) - whether to auto-create suggested Stack/Load LoRA nodes from the results, versus just producing a report.
• model (required) - used for key mapping and target grouping, same as the Optimizer needs it.
• lora_stack (required) - the LoRAs to analyze.
• clip (optional) - include it if you want text-encoder LoRA keys factored into the analysis.

Outputs: report - a STRING, wire it to a Show Text node for the readable breakdown. compatibility_map - an IMAGE, a visual rendering of how the LoRAs relate rather than a wall of numbers.

How to install it

ComfyUI Manager: search "LoRA Optimizer", install, restart. Manually:

cd ComfyUI/custom_nodes/
git clone https://github.com/ethanfel/ComfyUI-LoRA-Optimizer.git

Restart - it's under loaders. No downloads; it analyzes files you already have.

Common issues & troubleshooting

Nothing happens when you run the workflow. Check enabled first - it's off by default specifically so this node doesn't add overhead to workflows where it's just sitting on the canvas unused. This is the single most common reason this node appears to do nothing.

Extra nodes appeared on your canvas that you didn't add. That's create_nodes doing its job - it's on by default and will build LoRA Stack / Load LoRA nodes from the analysis results. If you just want the report without the pack rearranging your graph, turn create_nodes off.

You want an actual merge, not just an analysis. This node stops at diagnosis on purpose - it never produces a MODEL/CLIP output. Once you know a combination looks promising, take the same lora_stack to LoRA Optimizer or LoRA AutoTuner to actually merge it.