AKASHIC (JOV) 📓

AKASHIC (JOV) 📓 is a "show me what this actually is" node. It takes any input - image, mask, text, latent, number, whatever Jovimetrix can hold - and renders it for you to read: images get previewed, text gets printed, and everything else gets a human-readable description instead of a wall of tensor shape. In a pack full of *-typed ports that happily accept anything, that's not a luxury, it's how you figure out what's flowing through your graph.

The name is a bit much (the "akashic records" thing - fine, the author has a theme), but the function is dead practical. This is the Jovimetrix answer to the debugging problem every ComfyUI user eventually hits: you can't tell if a node is receiving a 512×512 RGBA image or a batch of 24 masks just by looking at the wire.

How it works

The node takes whatever you plug into it and builds a UI payload: base64-encoded image previews for tensors, text representations for numbers and strings, and structured output for the oddballs. Its own description spells it out - it accepts "images, text, and other types"; if no input is connected it returns an empty result. It's an output/display node (is_output_node), so its job is the panel, not a downstream connection.

What's nice is the type sniffing. Jovimetrix types its inputs loosely on purpose, and AKASHIC will tell you what it actually got: it decodes latents (samples dicts) to their tensor shape, recognizes Mixlab layer dicts, unwraps Jovimetrix vectors, and pretty-prints plain dictionaries as JSON. That last one is gold when you're dealing with a metadata dict or a workflow that passes structured data around.

Inputs and outputs

There are no fixed named inputs - you get dynamic slots for whatever you want to inspect, and no data outputs (it's a sink). If nothing is wired in, it returns an empty result rather than erroring, which makes it a safe node to drop into a half-built workflow.

Installing it

It ships with Jovimetrix:

• ComfyUI Manager - search "Jovimetrix" and install.
• Manual - git clone https://github.com/Amorano/Jovimetrix.git into ComfyUI/custom_nodes/, then pip install -r requirements.txt.

Same dependency story as the rest of the pack: numpy, OpenCV, Pillow, matplotlib, and the author's cozy_comfyui library from GitHub. No models. Needs ComfyUI 0.1.3+.

Using it in practice

Drop it at the end of a chain you're unsure about. If the preview looks wrong, the problem is upstream of that wire. If you see a tensor shape you didn't expect, you've caught the bug. It pairs well with the pack's IMAGE INFO node - AKASHIC for "what kind of thing is this," IMAGE INFO for "exact dimensions." One genuinely useful habit: leave one in a workflow you're actively editing, and remove it before you share the JSON, because it does add a little overhead on every run. It's not the flashiest node in Jovimetrix, but it's the one that turns "why is my output garbage?" into a question with an answer.