GLYPH

GLYPH is a codec-first structured text format and streaming substrate for AI and ML systems.

The core value is lower in the stack than “agent framework”:

• deterministic loose-mode canonicalization
• JSON bridges
• fingerprinting for state verification
• schema-driven packed, tabular, and patch encodings
• GS1 framing for multiplexed streams
• cross-language implementations in Go, Python, JavaScript, Rust, and C

Why This Repo Exists

JSON is a fine interchange format, but it is noisy for repeated structured data and weak as a codec substrate for long-running LLM workflows. GLYPH focuses on:

• Compact text: fewer tokens than JSON for many structured payloads
• Determinism: canonical output suitable for hashing and equality
• Patchability: explicit delta-friendly representations
• Streaming: frame-oriented transport and incremental validation
• Parity: the same semantics across multiple runtimes

If you are evaluating glyph, start from the codec and spec layer first. Higher-level agent patterns in this repo are examples, not the product center.

Install

Language	Package	Docs
Python	pip install glyph-py	Python README
Go	go get github.com/Neumenon/glyph	Go README
JavaScript / TypeScript	npm install cowrie-glyph	JS README
Rust	cargo add glyph-rs	Rust README
C	build from source	C README

Quick Example

import glyph

data = {"action": "search", "query": "glyph codec", "limit": 5}

text = glyph.json_to_glyph(data)
# {action=search limit=5 query="glyph codec"}

value = glyph.parse(text)
query = value.get("query").as_str()

fingerprint = glyph.fingerprint_loose(glyph.from_json(data))

Documentation Map

Start Here

• Quickstart
• Documentation Index

Authoritative Specs

• Loose Mode Spec
• GS1 Spec

API / Language Docs

• API Reference
• Python
• Go
• JavaScript / TypeScript
• Rust
• C

Example / Historical Material

• Agent Patterns — optional higher-level examples
• Research Reports — dated snapshots, not current API docs
• Archive — historical material

Repo Layout

glyph/
├── docs/                  authoritative specs, quickstart, index
├── go/                    Go implementation
├── py/                    Python implementation
├── js/                    JavaScript / TypeScript implementation
├── rust/                  Rust implementation
├── c/                     C implementation
└── tests/                 cross-implementation parity scripts

Current Positioning

The codec layer is the product:

• format contract
• canonicalization
• schema / packed / tabular / patch behavior
• streaming transport
• parity and correctness

The demo and agent-oriented material in this repo should be read as examples built on top of that substrate.