💪TB | Unpack Boolean

The TinyBreaker pack routes its settings around as one opaque genparams object. That's great for keeping the graph tidy, but what do you do when a standard ComfyUI node needs to know whether some flag is on? 💪TB | Unpack Boolean is the answer: it reaches into the bundle, pulls out one named boolean value, and hands it to the rest of your graph as a normal BOOLEAN.

Think of it as the translator between the pack's parameter system and the wider ecosystem. The pack speaks genparams; your Preview Image, your conditional node, your bypass logic speak plain values. This node sits in the middle.

The two inputs

• genparams - the bundle to read from.
• param_name - the dotted key you want, e.g. image.enable_upscaler. The tooltip is blunt about the behavior: if the name isn't found, it returns False. No error, no drama - just a clean fallback.

Output is a single BOOLEAN.

How it works and when you'd use it

Under the hood it's one line: genparams.get_bool(param_name, False). GenParams keys live in a nested, dotted namespace (image.*, denoising.base.*, user.*), so you need to pass the full path, not a shorthand. A real-world example: the Unified Prompt's --upscale on writes image.enable_upscaler, and if you're building a hybrid graph where a standard node needs to know whether upscaling will fire, this is how you read that flag.

Where it shines is conditional wiring - connect the boolean to something like an "if enabled, do X" switch and let the prompt text decide the graph's behavior. It's a small node, but it's the difference between a workflow that's rigid and one where a single --flag in a text box reroutes the pipeline. For a genuinely thin utility, that's a lot of leverage.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/martin-rizzo/ComfyUI-TinyBreaker

or search "tinybreaker" in ComfyUI Manager and restart. No Python dependencies - the pack ships zero requirements.txt. As with everything in this pack, the real requirement is the model files in models/checkpoints and models/clip, though this particular node technically works even before you've got them, since it never touches the model.

One gotcha worth knowing: the parameter name has to match the key exactly as the writing node stored it. Typos don't throw - they just silently return False - so if a flag reads as off when it should be on, check the spelling in param_name first. That silent fallback is the feature and the trap in one.