TOON

Token-Oriented Object Notation (TOON) compresses JSON-like payloads for LLM prompts. ToDiagram displays TOON content as diagrams so you can inspect model settings, chat messages, and tool schemas without scanning dense text.

Example

model: gpt-3.5-turbo-1106
response_format:
  type: json_object
messages[2]{role,content}:
  system,You are a helpful assistant that extracts data and returns it in JSON format.
  user,What is the weather like in Boston?
functions[1]:
  - name: get_current_weather
    description: Get the current weather in a given location
    parameters{type,properties,required}:
      type: object
      properties:
        location:
          type: string
          description: The city and state, e.g. San Francisco, CA
        unit:
          type: string
          enum[2]: celsius,fahrenheit
      required[1]: location
function_call: auto

What is TOON?

TOON is an open specification for storing JSON-style data with fewer tokens. It keeps nested structure with indentation while allowing inline tables, folded keys, and array annotations. Read the full spec at github.com/toon-format/spec.

The format shines when you need deterministic, compact prompts for LLMs yet still want a way to inspect the structure at a glance.

Working with TOON in ToDiagram

• Paste TOON data to see the structure laid out as nodes and edges, including model configuration, chat turns, and tool schemas.
• Collapse, search, or highlight nodes to verify prompt content before sending it to an LLM.
• Export the current diagram as PNG/SVG or share a cloud link when the diagram needs to be reviewed with teammates.

Converting to and from TOON

You can convert between any of the JSON-compatible formats supported by ToDiagram (JSON, YAML, XML, CSV) and TOON:

1. Paste TOON to view a diagram and optionally export the content back as JSON.
2. Paste JSON/YAML/XML/CSV data, visualize it, and download it as TOON if you need the compact representation.
3. Use the download button to keep both representations in sync for downstream tools.

This allows you to review prompts visually and still keep TOON payloads ready for model input.