LLM
JSON ↔ TOON Converter
Round-trip JSON into TOON outline notation with a single click and keep prompts token-lean.
JSON to TOON
Parse JSON safely and emit TOON with sorted keys for consistent diffs.
Validation happens in-browser—no payloads leave the page.
TOON output
products[2]{id,in_stock,price,product_name,tags}:
101,true,299.99,Wireless Noise Cancelling Headphones,[audio, bluetooth, premium]
102,true,145.5,Ergonomic Mechanical Keyboard,[peripherals, office, typing]TOON to JSON
Outline notation is parsed with the JSON-only schema to avoid surprises.
Two-space indents keep hierarchies obvious for humans and models.
JSON output
{
"products": [
{
"id": 101,
"in_stock": true,
"price": 299.99,
"product_name": "Wireless Noise Cancelling Headphones",
"tags": [
"audio",
"bluetooth",
"premium"
]
},
{
"id": 102,
"in_stock": true,
"price": 145.5,
"product_name": "Ergonomic Mechanical Keyboard",
"tags": [
"peripherals",
"office",
"typing"
]
}
]
}Why TOON?
- Token-lean structure: Dropping braces and commas typically trims prompt token count by 10–20%, leaving more space for actual content.
- Indent-first clarity: Tree depth is obvious to humans and models, which reduces hallucinated siblings when an LLM rewrites data.
- Type-disciplined parsing: The converter enforces JSON-safe scalars so TOON stays compatible with APIs and config files.
Why LLM teams like it
TOON is trending with LLM-heavy stacks because it reads like an outline, keeps deltas tiny for review, and round-trips cleanly to JSON for execution. It also plays well with constrained decoding strategies that expect deterministic indentation.
plan:
summary: "Keep output terse and structured"
steps:
- capture goals
- normalize inputs
- stream final JSON