Adaptive Token-Oriented Notation
The data serialization format designed for Large Language Models.
Up to 56% token reduction with zero data loss.
What makes ATON perfect for LLMs and AI applications
Dramatically lower your LLM API costs. Save thousands on GPT-4, Claude, and other models.
Explicit schema with type definitions. LLMs understand your data structure perfectly.
Easy to read, write, and debug. Both humans and AI can understand it instantly.
Reference entities directly with → syntax. Perfect for RAG systems and knowledge graphs.
Omit repetitive values automatically. @defaults reduces redundancy across records.
Perfect round-trip conversion. decode(encode(data)) === data, always.
Enterprise features for large-scale AI deployments
Fast, Balanced, Ultra, Adaptive - choose your speed/size tradeoff
WHERE, SELECT, ORDER BY - filter before encoding
Process unlimited data with O(1) memory
Complete error handling, validation, type safety
See how ATON transforms complex enterprise data
Retrieval-Augmented Generation system processing 10,000 queries per day with complete metadata.
{
"documents": [
{
"document_id": "doc_2024_001",
"filename": "Q4_Report.pdf",
"status": "processed",
"total_pages": 87,
"file_size_bytes": 2458624,
"processing_time_ms": 3450,
"confidence_score": 0.978,
"language": "en",
"document_type": "financial",
"classification": "confidential",
"compliance_tags": ["SOX", "GDPR"],
"embeddings_generated": true,
"embedding_model": "ada-002"
}
],
"chunks": [
{
"chunk_id": "ch_001",
"document_id": "doc_2024_001",
"page_number": 1,
"content": "Q4 revenue $145.7M...",
"confidence_score": 0.982,
"entities": ["revenue", "growth"],
"created_at": "2024-11-15T09:35:22Z"
}
/* ... 49 more chunks */
]
}
@schema[doc_id:str, filename:str, pages:int,
size:int, process_ms:int, confidence:float]
@defaults[status:"processed", language:"en",
embeddings:true, model:"ada-002"]
documents(1):
doc_2024_001, "Q4_Report.pdf", 87, 2458624,
3450, 0.978, "financial", "confidential",
["SOX","GDPR"]
@schema[chunk_id:str, doc_ref:ref, page:int,
content:str, confidence:float,
entities:arr, created:datetime]
chunks(50):
ch_001, ->documents[doc_2024_001], 1,
"Q4 revenue $145.7M...", 0.982,
["revenue","growth"],
2024-11-15T09:35:22Z
/* ... 49 more rows */
ATON vs JSON, and CSV/VSC
| Feature | JSON | CSV/VSC | ATON |
|---|---|---|---|
| Token Efficiency | Baseline | 71% better | 56% better |
| Type Safety | ✓ | ✗ | ✓ |
| Nested Support | ✓ | ✗ | ✓ |
| Relationships | ✗ | ✗ | ✓ Native |
| Human Readable | ✓ | ✗ | ✓ |
| LLM Optimized | ✗ | ✗ | ✓✓ |
| Schema Flexibility | ✓ | ✓ | ✓ |
| Default Values | ✗ | ✗ | ✓ |
Convert your JSON to ATON and see the savings