These reports are dated research snapshots.
They are useful for historical context, benchmarking notes, and earlier experiments, but they are not authoritative API or installation docs. Package names, examples, and positioning may have drifted since some of these reports were written.
- Codec performance (parsing, canonicalization, fingerprinting)
- Speed metrics across languages (Go, Python, JS)
- Token reduction measurements
- Key finding: Go achieves 2M+ ops/sec canonicalization
- Performance optimization techniques
- Memory usage analysis
- Codec tuning results
- Index of all benchmark runs
- Historical performance data
- Comparison across versions
- How LLMs handle GLYPH vs JSON
- Retrieval accuracy across formats
- Generation quality comparison
- Embedding analysis
- Key finding: Hybrid approach recommended (LLMs generate JSON, store as GLYPH)
STREAMING_VALIDATION_REPORT.md
- Streaming validation performance
- Token-by-token error detection
- Latency savings measurements
- Key finding: Detect errors as they stream and cancel immediately, saving tokens and time
- Tool calling patterns
- System prompt token reduction
- Validation accuracy
- Error recovery strategies
Documentation:
- Quickstart - Get started in 5 minutes
- Complete Guide - Features and patterns
- Specifications - Technical specs
Code:
- Python - Python implementation
- Go - Go implementation
- JavaScript - JS implementation
Questions? Open an issue or check discussions.