NOBJ aims to provide a structured data format that is closer to natural language while retaining the simplicity and ease of parsing of JSON. The goal is to allow LLMs to output higher-quality results that can be directly parsed by programs without the need for secondary conversion to JSON.
- Readability: Uses words and structures closer to everyday language, making it easier for LLMs to read and understand.
- Maintainability: More intuitive design, making data structures easier to understand and modify.
- Cross-Language Usability: Natural language elements can more easily bridge language barriers for better understanding.
Current limitations:
- Supports only flat simple objects or arrays.
- Lacks hierarchical structure, unable to represent nested objects or arrays.
- Does not support optional fields.
- Requires prompt engineering to ensure LLM outputs in the correct order and format.
NOBJ Representation of a Simple Object:
name: John Doe
age: 30
city: New York
# Array values
tags: developer, engineerNOBJ Representation of a Simple Array:
name: John Doe
age: 30
city: New York
name: Mike Doe
age: 20
city: New YorkNotes:
NOBJis still under development, and specific syntax rules and parsing methods will be further refined in the future.- String values do not include newline characters.
- Array values are comma-separated.
- If a
keycontains spaces, the spaces in the returned JSON object'skeywill be replaced with underscores (_).