AXON Documentation
Complete guide to using AXON for token-efficient LLM interactions
What You'll Learn
- What AXON is and how it reduces LLM token costs by 60-95%
- Key features including 13 validated types and 5 compression algorithms
- When to use AXON versus JSON or CSV
What is AXON?
AXON (Adaptive eXchange Oriented Notation) is a data serialization format designed specifically for Large Language Model interactions. It achieves 60-95% token reduction compared to JSON while maintaining type safety, validation, and the flexibility to handle complex nested structures.
Unlike traditional formats optimized for human readability or machine parsing, AXON optimizes for the token economy of LLM APIs, making your AI applications more cost-effective and performant by reducing context rot.
Key Insight: AXON combines CSV's compactness with JSON's flexibility. For typical datasets, it matches or beats CSV while adding type annotations and validation that neither JSON nor CSV provide.
Key Features
- 60-95% token reduction versus JSON
- JSON & XML input support - convert from both formats seamlessly
- 13 validated types including u8, i32, f64, bool, iso8601, uuid, enum
- 5 compression algorithms automatically applied based on data patterns
- 6 adaptive modes for different data structures
- Schema system with inheritance and validation
- Query hints to optimize LLM understanding
- 93.51% test coverage with 342 passing tests
When to Use AXON
Choose AXON when you need:
- Token efficiency for LLM API calls
- Type validation before sending data to expensive APIs
- Compression on patterned or structured data
- Nested object support that CSV cannot provide
Note: JSON remains appropriate for simple, unstructured data where human readability matters more than token count. AXON excels at structured, tabular, or repetitive data.