Question 1

What is TOON format?

Accepted Answer

TOON (Token-Oriented Object Notation) is a compact data serialization format designed to minimize token usage in language model applications. It maintains full semantic compatibility with JSON while using fewer characters and tokens.

Question 2

How much can I save with TOON?

Accepted Answer

Token savings vary depending on your JSON structure, but typically range from 30% to 50% reduction compared to standard JSON. The exact savings are displayed in real-time when you convert your data.

Question 3

Is TOON compatible with JSON?

Accepted Answer

Yes, TOON maintains full semantic compatibility with JSON. You can convert TOON back to JSON using the decode function from the @toon-format/toon library. The data structure and values remain identical.

Question 4

How are tokens calculated?

Accepted Answer

We use tiktoken with the GPT-4 encoding model to calculate token counts. This provides accurate token counting that matches how OpenAI and other LLM providers count tokens.

Question 5

Is my data secure?

Accepted Answer

Yes, all conversions happen entirely in your browser. Your JSON data never leaves your device and is not sent to any server. This ensures complete privacy and security.

Question 6

Can I use TOON with any LLM?

Accepted Answer

TOON format can be used with any language model that accepts text input. However, the model needs to understand the TOON format. Some models may require you to decode TOON back to JSON first.

Question 7

What file format does the download use?

Accepted Answer

The downloaded file uses the .toon extension and contains plain text in TOON format. You can open it with any text editor and use it directly in your applications.

Question 8

Does the converter work offline?

Accepted Answer

Yes, once the page is loaded, all conversion happens client-side in your browser. You can use the converter offline after the initial page load.

Question 9

Can I convert large JSON files?

Accepted Answer

The converter can handle reasonably large JSON files, but very large files (several MB) may impact browser performance. For best results, we recommend files under 1MB.

Question 10

How accurate is the token counting?

Accepted Answer

We use tiktoken with the GPT-4 encoding model, which provides highly accurate token counting. The counts match how OpenAI counts tokens for GPT models, ensuring reliable estimates.

Frequently Asked Questions