It also enforces the correct structure for the different types like "function" and "object", so you can't make syntax or structural errors. Cursor didn't use the OpenAI format when I auto-completed a {"type": "function"} schema, but this isn't a problem using my tool.
Why use this over Pydantic? It's a lot more visual, and you can reorder properties and change "required" properties without wrangling with text, so it's easier when you're in the prototyping stage.
The source is at https://github.com/anusarati/json-schema-builder.
What do you all think?