Object Record Table (ORT)

ORT is a CSV-like structured data format with native support for objects and arrays.
Unlike inefficient human-readable data structures, ORT efficiently represents data, making it ideal for large language model (LLM) token optimization and structured data interchange.

Tip

ORT was inspired by Token-Oriented Object Notation (TOON) and was created to contribute to the open-source ecosystem and the advancement of LLMs.
Thank you, TOON.

Table of Contents

• Why ORT
• When Not to Use ORT
• Command-Line Tools
• Installation
• Package Links
• Specification
• License

Why ORT

Standard JSON and YAML prioritize human readability, but their structural characteristics introduce significant computational overhead when passing data to large language models. ORT addresses this by optimizing data representation for token efficiency.

JSON

{
  "users": [
    {
      "id": 1,
      "profile": {
        "name": "John Doe",
        "age": 30,
        "address": {
          "city": "New York",
          "country": "USA"
        }
      }
    },
    {
      "id": 2,
      "profile": {
        "name": "Jane Smith",
        "age": 25,
        "address": {
          "city": "London",
          "country": "UK"
        }
      }
    }
  ]
}

398 Characters / 118 Tokens

Compact-JSON

{"users":[{"id":1,"profile":{"name":"John Doe","age":30,"address":{"city":"New York","country":"USA"}}},{"id":2,"profile":{"name":"Jane Smith","age":25,"address":{"city":"London","country":"UK"}}}]}

198 Characters / 58 Tokens

ORT

users:id,profile(name,age,address(city,country)):
1,(John Doe,30,(New York,USA))
2,(Jane Smith,25,(London,UK))

110 Characters / 35 Tokens

When Not to Use ORT

ORT is not always the optimal choice. Consider using JSON in the following scenarios:

• Non-uniform data structures: ORT operates efficiently and reliably with uniform data structures. For heterogeneous data, JSON is more efficient and prevents data loss.
• Direct application data files: ORT is not optimized for direct input/output operations. Since ORT specializes in data transformation and processing, using JSON for application data files offers better compatibility and stability.

Command-Line Tools

The Rust implementation provides CLI utilities for converting between ORT and JSON formats:

Installation

cargo install ort-rs

Usage

ort2json <input_file>
json2ort <input_file>

# Or with output file specification

ort2json <input_file> -o <output_file>
json2ort <input_file> -o <output_file>

Installation & Usage

PLAYGROUND

ORT Playground

Rust Library

Add to your Cargo.toml:

[dependencies]
ort-rs = "1.1.0"

TypeScript Library

npm i ort-ts

Python Library

pip install ort-python

Package Links

• Crates.io: https://crates.io/crates/ort-rs
• PyPI: https://pypi.org/project/ort-python/
• npm: https://www.npmjs.com/package/ort-ts

Specification

To view the ORT specification sheet, go here.

License

This project is licensed under the MIT License. See the LICENSE file for details.