toon-syntax

You can reduce token usage in LLM prompts by encoding JSON data into a compact notation using indentation, explicit array lengths, and minimal quoting to save tokens. This approach replaces bulky braces and brackets with clear structural hierarchy.

The best way to serialize structured data for token-constrained AI workflows is using a compact, token-efficient notation. It converts JSON-like data by applying indentation, defining objects and arrays with clear rules, and supporting optional key folding for compact paths.

TOON supports flexible delimiters for data interchange, allowing you to use commas, tabs, or pipes. This delimiter flexibility helps optimize token efficiency across different contexts and structured data encoding environments.

You can encode arrays compactly to save tokens by using explicit [N] array length headers instead of traditional brackets. This notation also supports tabular arrays, dramatically reducing token usage while preserving data readability.

You can use minimal quoting rules for JSON data encoding to save tokens while preserving readability. By applying indentation-based syntax and only quoting where strictly necessary, you can effectively interchange structured data under strict size constraints.

Limitations of this indentation-based data encoding include relying heavily on consistent whitespace for structural hierarchy instead of explicit braces. You should not use this approach if your downstream parser cannot strictly preserve indentation or tab characters.