tron-format-py

A Python library for the TRON data serialization format — a compact JSON superset designed for LLM token efficiency.

---

TRON (Token Reduced Object Notation) extends JSON with schema-based class definitions, dramatically reducing token count when serializing structured data for LLM APIs. tron-format-py provides a complete Python implementation of the TRON specification.

Installation

pip install tron-format-py

Requirements: Python 3.10+

Quick Start

fromtron_format_pyimportTRON# Serialize to TRONvalue= [{"name": "Alice", "role": "admin"}, {"name": "Bob", "role": "user"}]
tron=TRON.stringify(value)
print(tron)
# class A: name,role## [A("Alice","admin"),A("Bob","user")]# Parse back to Pythonparsed=TRON.parse(tron)
assertparsed==value

Features

Feature	Description
Schema-based encoding	Repeated object structures become class instantiations, eliminating key repetition
Named arguments	Assign values by property name for improved readability
Class inheritance	Extend existing classes to model hierarchical data
Pretty printing	Optional indented output with configurable indentation
Full spec compliance	100% test coverage across all TRON format features
Round-trip safe	parse(stringify(data)) == data for all supported types

API Reference

TRON.stringify(value, indent=None)

Convert a Python object to a TRON format string.

Parameter	Type	Default	Description
value	Any	—	The Python object to serialize
indent	int | None	None	Number of spaces for indentation. None or 0 produces compact output

Returns:str — A TRON format string

Example:

# Compact output (default)TRON.stringify({"x": 1, "y": 2})
# '{"x":1,"y":2}'# Pretty outputTRON.stringify([{"x": 1}, {"x": 2}], indent=2)
# """class A: x## [# A(1),# A(2)# ]"""

TRON.parse(text)

Parse a TRON format string into a Python object.

Parameter	Type	Default	Description
text	str	—	A TRON format string

Returns:Any — The parsed Python object

Example:

TRON.parse('class Point: x, y\nPoint(10, 20)')
# {"x": 10, "y": 20}

Examples

Class Instantiation

Objects can be instantiated using defined classes, mapping positional arguments to class properties:

tron='''class Order: index, items, totalclass Product: index, name, price, quantityOrder("ord-123", [ Product(1, "Widget", 19.99, 2), Product(2, "Gadget", 29.99, 1)], 109.96)'''TRON.parse(tron)
# {# "index": "ord-123",# "items": [# {"index": 1, "name": "Widget", "price": 19.99, "quantity": 2},# {"index": 2, "name": "Gadget", "price": 29.99, "quantity": 1}# ],# "total": 109.96# }

Named Arguments

TRON.parse('class User: name, email\nUser(name="alice", email="alice@example.com")')
# {"name": "alice", "email": "alice@example.com"}

Class Inheritance

TRON.parse('''class Base: idclass Extended(Base): name, valueExtended(1, "test", 42)''')
# {"id": 1, "name": "test", "value": 42}

Supported Types

TRON Type	Python Type
null	None
true / false	True / False
123 / 12.34	int / float
"string"	str
[a, b, c]	list
{a: 1, b: 2}	dict
Class(a, b)	dict (keys from class definition)

Links

• Specification:tron-format.github.io
• Repository:github.com/tron-format/tron-python
• Issues:github.com/tron-format/tron-python/issues
• PyPI:pypi.org/project/tron-format-py

License

MIT License. Copyright (c) 2025-2026 Tim Huang. See LICENSE for details.