JsonValidate (Yogurt Nodes)

When an LLM or a web service hands you JSON, the two questions are: is it valid and is it the shape I expected? JsonValidate answers both. Feed it any data, tell it what type you expect (and which keys have to be there), and it returns is_valid - a BOOLEAN - plus an error_message string that tells you what was wrong when it isn't.

It's a logic node in the YogurtNodes pack (yogurt7771/ComfyUI-YogurtNodes), and it's the quality gate for the pack's JSON workflow. Think of it as the graph version of a schema check: light, explicit, and it gives you a reason instead of just a thumbs-down.

How it works

Two checks, both simple under the hood. First, check_type (default any) verifies the data is the expected JSON type - object, array, string, number, boolean, null, or any for "don't care". Second, required_keys - a comma-separated list - checks that every listed key is present when the data is an object (it's ignored for arrays and scalars). If both pass, is_valid is True and error_message is empty; if either fails, you get False and a description of which check tripped.

The two outputs are designed to be used together: branch on is_valid, and if you're in the failure branch, log or show error_message so you (or your workflow) know why instead of just knowing it failed.

Inputs that matter

• data - anything you want checked.
• check_type - the expected type. The one beginners skip and then wonder why a list passed their "object" check.
• required_keys - comma-separated, e.g. id,title,url. Only meaningful for objects, but it's the check that catches "the API changed its schema."

Outputs: is_valid (BOOLEAN) and error_message (STRING).

Install

Pack-wide routine - ComfyUI Manager (search "YogurtNodes") or:

cd ComfyUI/custom_nodes
git clone https://github.com/yogurt7771/ComfyUI-YogurtNodes.git
cd ComfyUI-YogurtNodes
pip install -r requirements.txt

Restart, find it under Yogurt Nodes / Logic. No extra dependencies.

Troubleshooting

• "It's a valid JSON string but validation fails." Validation works on objects, not text. A JSON string is a string type until you Parse it. Stringify's output isn't valid-here - parse first, validate the parsed object.
• "It passed but it's clearly wrong." required_keys checks presence, not value quality. {"id": null} passes if id is listed. For "present and non-null", combine with GetPath and a null check.
• Type gotchas. JSON null is a distinct type - a null check only passes for actual nulls. And be careful: in JSON, number includes integers and floats; there's no separate integer type here.
• "Why did my whole workflow die?" If you're validating at the input and then not branching on the result, you've built the check but ignored it. Wire is_valid into a Switch so a bad payload takes a fallback path (or a clear error output) instead of silently failing deeper in the graph.

The workflow worth building: parse → validate with required keys → branch. One node turns "trust me" JSON into "I checked" JSON, and the error_message output means the failure message lands in your workflow where you can act on it, not in a log you'll find an hour later.