Petty Paint To Json

The petty-paint setup runs on JSON. The iPad app writes a project file, nodes talk to each other with JSON blobs, and if you're building a workflow around it you will eventually be staring at a string that is structured data but refuses to act like it. Petty Paint To Json is the decoder ring: it takes JSON text on one side and hands you the parsed Python object on the other - a dict, a list, whatever was in there - so the pack's PettyPaintJsonRead, PettyPaintJsonMap, and PettyPaintExec nodes (or anything else that accepts any) can actually traverse it.

How it works

Under the hood it's one call: json.loads(text). If the input is already not a string - say you're chaining two JSON nodes together - it passes the value through untouched instead of double-parsing. The output is typed as * (any), because the result could be a dictionary, a list, a number, or a nested mix of all three, and ComfyUI can't know until it runs.

The inputs and outputs that matter

• text - the JSON string to parse. It's forceInput, so you wire it in from another node rather than typing into the box.

• data - the parsed object. This is what you feed into the pack's JSON-walking nodes or into any any-typed input downstream.

One thing to notice: this node does no validation beyond "is this valid JSON." If the string isn't valid JSON, json.loads raises, and your whole queue errors out with a traceback pointing here. That's the most common way people meet this node. If you're parsing data that might be empty or half-written, either guard it upstream or accept that a malformed blob will kill the run - there's no graceful-fallback mode.

Installing it

It's part of the petty-paint pack, so you get it with everything else. ComfyUI Manager search "Petty Paint", or:

cd ComfyUI/custom_nodes
git clone https://github.com/mephisto83/petty-paint-comfyui-node

restart, done. No models, no heavy installs - the pack's only pinned requirement (Flask) isn't even imported by the current source.

When it bites you

Beyond invalid JSON, the classic trap is feeding it a plain human-readable prompt instead of JSON - it'll throw immediately, which is at least an honest failure. And keep in mind the output is a live Python object, not a string anymore; if you want to log it or put it in a filename, you'll need to stringify it first (the pack's PettyPaintConvert covers that). Use it to bridge the gap between "text" and "structured data" and it does exactly one job, reliably.