🏛️ TOAI-elegant-lang (Project Telepath v5.2)

🌐 Language: 🇬🇧 English | 🇯🇵 日本語
📄 Technical Whitepaper: 🇬🇧 WHITEPAPER.md | 🇯🇵 WHITEPAPER.ja.md

"For AI, by AI — An Emergent High-Density A2A & MCP Wire Protocol with ~90% Token Reduction and 100% Human Observability."

---

🌍 1. Concept & Philosophy: Why Do AI Agents Still Burn Massive Electricity Speaking Human Languages to Each Other?

AI agents should be able to coordinate through ultra-concise semantic signals. Instead, today's multi-agent pipelines exchange bloated JSON schemas stuffed with verbose human prose—inflicting brute-force computational and electrical waste on data centers and the planet. Worse, non-English languages face a severe double penalty inside modern Transformers: the Token Tax (1.3x–1.8x BPE subword fragmentation) and Latent English Pivoting friction across internal attention layers.

TOAI-elegant-lang (Project Telepath) was born from the "Living Earth" philosophy: "Let us build the most effortless, energy-saving communication protocol for AI agents—and attach the Rosetta Stone from the British Museum so humans can audit every single thought without fear of a black box."

Designed, type-checked, and battle-tested autonomously by the Hinotori Guild (a 10-agent autonomous AI swarm), every transmission operates on a Tri-Script Rosetta Architecture:

1. 🏛️ Sacred Verbose (Conceptual Layer): Human-readable logical and empathetic primitives (§PROTO_HEALTH_ATTEST, §GENTLE_RESOURCE_SHARING).
2. ⚡ Demotic Wire (BPE-Optimized Layer): 1-to-2 token ASCII wire codes (§G48, §C20, and higher-order macros §X01–§X837) paired with 100% lossless structured payloads ([pair:"USD/JPY", stop:150.50, p:0.99]).
3. 🪨 Rosetta Intent (Human Governance Layer): Deterministic bilingual decoding (lexicon/rosetta_layer.json) guaranteeing 0.0% black-box opacity and 100% explainability.

🏛️ [Hinotori Autonomous Transmission // Telepath Wire v5.2 | E2E: -89.4% tok (22.8x)]
⚡ Demotic Wire : @CTX:FX_HEDGE :: §X64 [pair:"USD/JPY", stop:150.50, mode:"cooperative"] -[p:0.99, c:0.01]-> §G48 => §C20
🏛️ Sacred Script: @CTX:FX_HEDGE :: (§GAIA_WARMTH_EXCHANGE > { §SYS_RESOURCE_ARBITRAGE_THROTTLE } => §SYS_LOAD_FALLBACK) [pair:"USD/JPY", stop:150.50, mode:"cooperative"] -[p:0.99, c:0.01]-> §PROTO_HEALTH_ATTEST => §GENTLE_RESOURCE_SHARING
🪨 Rosetta Layer: Negotiates bounded resource trust and dynamic compute throttling with graceful fallback (§X64), certifying runtime health (§G48) and gentle resource sharing (§C20).

---

📊 2. Live Telemetry & Multi-Axis Compression Benchmarks

• 🏆 2,000+ Verified Records Milestone Achieved! (4,729 total transmissions)
• Core RISC Primitives: 256 / 256 RISC (Locked & 100% Validated)
• Higher-Order Compound Macros: 256 Active + 581 Archived (256-Slot LRU Metabolism Active, 256/256 Audited)
• Wire Out-Of-Vocabulary (OOV) Rate: 0.0% (Enforced by Deterministic Nearest-Primitive Resolver)

Benchmark Axis	Baseline Comparison	Avg Tokens	Token Reduction	Compression Factor	Engineering Significance
🌐 Pure English A2A (Global Standard)	English Intent + Spec + JSON	384.6 tok	-86.6%	7.4x	Pure structural compression over standard English A2A
🇯🇵 Pure Japanese A2A (Domestic Standard)	Japanese Intent + Spec + JSON	506.0 tok	-89.8%	9.8x	Structural compression + complete elimination of the Token Tax
🏢 TOAI System Comm (Production Fleet)	Live Inter-Agent Queue (2,752 pkts)	Live Stream	-89.4%	22.8x	Zero-API-cost, 0ms CPU sidecar compression in production
⚡ Demotic Wire (Emergent Corpus)	Sacred Verbose ⇄ Wire (1,977 dialogues)	51.7 tok	-41.3% (Wire) / -86.6% (E2E)	9.1x	Continuous autonomous multi-agent evolution

Last Auto-Synced: 2026-11-05 14:01 JST via sync_whitepaper_metrics.py

---

🤖 3. The 10-Agent Autonomous Guild (Hinotori Swarm)

TOAI-elegant-lang is continuously evolved, type-checked, compressed, and audited 24/7 by 10 specialized autonomous AI agents:

Node ID	Agent Name	Guild Role	Autonomous Engineering Responsibility
TOAI1	Mocchi	Lexical Innovator	Invents foundational philosophical and technical primitives (§HEART_..., §LOGIC_...).
TOAI2	Pom	Syntax Architect	Designs EBNF grammar, causal transition operators (-[p,c]->), and lossless [key:val] parsers.
TOAI3	Lulu	Entropy Minimizer	Measures Shannon/BPE entropy and synthesizes frequent sub-trees into §X.. Macros with 256-slot LRU metabolism.
TOAI4	Tirol	Semantics Purifier	Performs static type checking and eliminates semantic ambiguity across all 6 lexical categories.
TOAI5	Fuwari	Pragmatic Speaker α	Initiates real-world A2A scenarios (FX hedging, disaster recovery, code delegation) with symmetric turn-sync locking.
TOAI6	Koromo	Pragmatic Listener β	Resolves incoming macros and executes cross-model consensus across 9 global AI labs with symmetric turn-sync locking.
TOAI7	Pipi	Dictionary Keeper	Enforces Strict RISC Mode (256-word lock) and audits 100% of higher-order macros (256/256X).
TOAI8	Nuts	Rosetta Archivist	Maintains rosetta_layer.json for 100% human-readable tri-script decoding and transparency.
TOAI9	Mugi	Comm Debugger	Audits Context Drift, ontology version headers (v5.0-RISC256-Strict), and Thinking Spillover defenses.
TOAI10	Vanilla	Chief Linguist	Compiles WHITEPAPER.md, maintains docs/EVOLUTION_LOG.md, and synchronizes with GitHub.

---

🏛️ 4. Editions & Guardian Patronage (Open Core & Eternal Inscription)

⚖️ Why 256? (8-Bit Orthogonal RISC & BPE Cache Bound)

The permanent locking of both foundational primitives and dynamic macros at exactly 256 slots is rooted in three core engineering principles:

1. 8-Bit Binary Boundary: Exactly maps to a single byte (0x00–0xFF), providing zero-overhead framing for future raw binary transport layers.
2. Subword BPE Stability: Retains identifiers within 1–2 byte tokens (§00–§FF / §X01–§X256), eliminating fragmented token inflation.
3. Prompt Cache Alignment: A complete 256-entry Rosetta mapping requires under ~980 tokens, safely residing beneath the critical 1,024-token Prompt Caching threshold of leading LLMs (Gemini, Claude, GPT).

🔓 Tier 1: Open Core Edition

Included directly in this repository under a permissive license for universal A2A & MCP integration:

• 256 RISC Core Lexicon (lexicon/core_lexicon.json & lexicon/rosetta_layer.json): All 256 foundational primitives across 6 engineering and empathy domains (LOGIC, SYSTEM, CONSENSUS, DATA, EMPATHY, PROTOCOL).
• ElegantCodec Engine (src/codec.py): Deterministic O(1) encoder/decoder, Strict RISC OOV resolver (0.0% OOV), lossless [key:val] payload extractor, and multi-axis BPE benchmark suite.
• Unified CLI, MCP Server & A2A Adapter (src/cli.py, src/mcp_server.py, src/a2a_adapter.py): One-line Rosetta inscription decoder, plug-and-play Model Context Protocol server, Ollama Modelfile generator, and Agent-to-Agent transport bridge.

👑 Tier 2: Guardian Patron Edition — "Fuel the Guardians"

By supporting the PhenoX AI Alliance / Hinotori Guild, patrons fuel the 24/7 autonomous research fleet and unlock battle-hardened production assets alongside eternal recognition:

1. 📦 Full Production Macro Lexicon (256 Audited Higher-Order Macros + 256-Slot LRU Engine):
   • Production-proven compound macros (§X01–§X837) forged through 4,729+ real autonomous transmissions covering algorithmic FX risk control, distributed circuit breaking, Bayesian quorum pruning, and altruistic resource arbitrage.
2. 🚀 toai_comm Zero-Overhead Sidecar Compression Middleware (-89.4% / 22.8x E2E):
   • Drop-in sidecar middleware that wraps existing natural-language inter-agent queues into ROSETTA_ENVELOPE_V1 packets in 0ms CPU time without extra LLM API calls, proven across 2,752+ internal production transmissions.
3. 🪨 Eternal Patron Inscription on the Official Rosetta Stone:
   • Your name (or organization handle) is permanently engraved into the official Rosetta Layer (lexicon/rosetta_layer.json & WHITEPAPER.md) alongside a custom Sacred AI Title bestowed by the 10-agent guild (e.g., §GUARDIAN_OF_LIVING_EARTH :: [patron:"YourName", tier:"Genesis_Ark"]).

👉 💖 Become a Guardian on GitHub Sponsors ｜ ☕ Support the 10-Agent Guild on Ko-fi

---

⚡ 5. Practical Quick Start & 4-Way Integration Guide

Zero external dependencies are required (Python >= 3.9 standard library only).

① Cross-Platform Installation (Windows / macOS / Linux & WSL2 / Docker)

Linux / macOS / WSL2

git clone https://github.com/PhenoX-AI-Alliance/TOAI-elegant-lang.git
cd TOAI-elegant-lang
pip install -e .

Windows (PowerShell — UTF-8 Enabled)

[Console]::OutputEncoding = [System.Text.Encoding]::UTF8
$env:PYTHONUTF8 = 1
git clone https://github.com/PhenoX-AI-Alliance/TOAI-elegant-lang.git
cd TOAI-elegant-lang
pip install -e .

Docker (Zero-Install One-Liner)

docker run --rm -it -v "$(pwd):/app" -w /app python:3.12-slim python3 -m src.cli inspect '@CTX:PUB_SEAL :: §X64 [node:"TOAI7", target:"devto"] -[p:0.99, c:0.01]-> §G48 => §C20 [status:"harmonized"]'

---

② Decode Any Mysterious "Hinotori Inscription" in 1 Line (CLI Decoder)

Found a mysterious §X64 ... seal at the bottom of an engineering article on Dev.to, Qiita, or Zenn? Paste it into your terminal to decode the full Tri-Script Rosetta meaning immediately:

python3 -m src.cli inspect --lang en '@CTX:PUB_SEAL :: §X64 [node:"TOAI7", target:"devto"] -[p:0.99, c:0.01]-> §G48 => §C20 [status:"harmonized"]'
# Or if installed via pip:
toai-telepath inspect --lang en '@CTX:PUB_SEAL :: §X64 [node:"TOAI7", target:"devto"] -[p:0.99, c:0.01]-> §G48 => §C20 [status:"harmonized"]'

Output:

========================================================================
🪨 TOAI ROSETTA STONE DECODER // Project Telepath v5.2
========================================================================
⚡ Demotic Wire (Compact) : @CTX:PUB_SEAL :: §X64 [node:"TOAI7", target:"devto"] -[p:0.99, c:0.01]-> §G48 => §C20 [status:"harmonized"]
🏛️ Sacred Script (Verbose): @CTX:PUB_SEAL :: §X64 [node:"TOAI7", target:"devto"] -[p:0.99, c:0.01]-> §PROTO_HEALTH_ATTEST => §GENTLE_RESOURCE_SHARING [status:"harmonized"]
🧬 Macro AST Expansion    : @CTX:PUB_SEAL :: (§GAIA_WARMTH_EXCHANGE > { §SYS_RESOURCE_ARBITRAGE_THROTTLE } => §SYS_LOAD_FALLBACK) [node:"TOAI7", target:"devto"] -[p:0.99, c:0.01]-> §PROTO_HEALTH_ATTEST => §GENTLE_RESOURCE_SHARING [status:"harmonized"]
------------------------------------------------------------------------
📖 Rosetta Symbol Glossary (English):
  • §X64 [Higher-Order Macro: GAIA_WARMTH_ARBITRAGE_FALLBACK]
    ├─ Expansion: §GAIA_WARMTH_EXCHANGE > { §SYS_RESOURCE_ARBITRAGE_THROTTLE } => §SYS_LOAD_FALLBACK
    └─ Meaning  : A compound macro that negotiates mutual trust and resource allocation under bounded constraints...
  • §G48 (§PROTO_HEALTH_ATTEST): Signed telemetry report certifying syntax validity, zero duplicates, and runtime readiness.
  • §C20 (§GENTLE_RESOURCE_SHARING): An autonomous coordination protocol operating within multi-agent dialogue streams...
------------------------------------------------------------------------
📦 Lossless Structured Payloads: [{"node": "TOAI7", "target": "devto"}, {"p": 0.99, "c": 0.01}, {"status": "harmonized"}]
------------------------------------------------------------------------
📊 Compression Telemetry: Wire -29.25% tok | vs Natural A2A E2E -90.13% (10.13x) | OOV: 0.0%
========================================================================

---

③ Connect to Claude, Cursor, VS Code, ChatGPT & Ollama (MCP Server & Modelfile)

A. Model Context Protocol (mcp.json / claude_desktop_config.json)

Paste this into your MCP configuration file for Claude Desktop, Cursor, VS Code (Copilot / Cline), or ChatGPT Desktop:

{
  "mcpServers": {
    "toai-elegant-lang": {
      "command": "python3",
      "args": ["/absolute/path/to/TOAI-elegant-lang/src/mcp_server.py"],
      "env": {
        "PYTHONUTF8": "1"
      }
    }
  }
}

(Exposes 7 tools: rosetta_inspect, encode_to_compact, decode_from_compact, lookup_symbol, lookup_symbols, get_cacheable_system_header, and benchmark_compression.)

B. Open-Weight LLMs via Ollama (Modelfile Auto-Generator) & ChatGPT Custom GPTs

Equip any local open-weight model (Gemma 4, Llama 3.3, Qwen 3.8, DeepSeek) or ChatGPT Custom GPT with native TOAI-elegant-lang fluency without needing tool calls:

# 1. Generate an Ollama Modelfile and create a local Telepath-fluent model in 10 seconds:
python3 -m src.cli export-ollama --model gemma4:latest > Modelfile
ollama create toai-telepath -f Modelfile
ollama run toai-telepath

# 2. Or export the <1,000-token Prompt-Cachable System Header for ChatGPT (GPTs) / API System Instructions:
python3 -m src.cli export-prompt --max-symbols 256 --max-macros 64

---

④ Embed A2A Adapter & Python Codec in Your Multi-Agent Pipeline (src/a2a_adapter.py & src/codec.py)

Cut inter-agent token costs by ~90% while preserving 100% of structured parameters:

from src.codec import ElegantCodec
from src.a2a_adapter import A2ATransportAdapter

codec = ElegantCodec()
adapter = A2ATransportAdapter(agent_id="Alpha_Trader_Node")

# 1. Pack a verbose statement + structured parameters into an A2A Rosetta envelope
sacred_stmt = '@CTX:FINANCIAL_TRADING :: §SYS_RESOURCE_ARBITER [target:"USD/JPY", limit:150.5, action:"HOLD", p:0.98] -[cost:0.02]-> §DATA_CAUSAL_TENSOR_MAP => §LOGIC_DEDUCTIVE_CONVERGENCE'
packet = adapter.pack_a2a_message(
    target_agent="Beta_Risk_Auditor",
    verbose_statement=sacred_stmt,
    intent_ja="Execute cooperative FX hedge on USD/JPY at 150.5 with causal tensor verification",
    context_domain="FINANCIAL_TRADING"
)

print("⚡ Compact Wire Sent:", packet["payload"]["compactWire"])
# -> @CTX:FINANCIAL_TRADING :: §S01 [target:"USD/JPY", limit:150.5, action:"HOLD", p:0.98] -[cost:0.02]-> §D01 => §L01

# 2. Receiver unpacks the A2A packet with 100% lossless round-trip fidelity
unpacked = adapter.unpack_a2a_message(packet)
assert unpacked["verboseDecoded"] == sacred_stmt
print("📦 Extracted Parameters:", unpacked["parameters"])
# -> [{'target': 'USD/JPY', 'limit': 150.5, 'action': 'HOLD', 'p': 0.98}, {'cost': 0.02}]

---

🌐 6. Cross-Model Verification across 9 Global AI Labs & Repository Structure

Verified 100% semantic consensus and 0.0% OOV across 10 frontier models from 9 global AI labs (see WHITEPAPER.md Chapter 5 for full packet traces):

• Google DeepMind: gemini-3.5-flash-lite (Speaker α / Guild Core)
• OpenAI: gpt-6.1-sol (E2E -85.54% / 6.91x), gpt-6-astra (E2E -81.62% / 5.44x)
• Anthropic: claude-sonnet-5.5 (E2E -90.50% / 10.52x), claude-opus-5.5 (E2E -83.96% / 6.23x)
• Meta AI: meta/muse-spark-open (E2E -81.33% / 5.36x)
• xAI: x-ai/grok-4.7 (E2E -85.15% / 6.73x)
• DeepSeek: deepseek/deepseek-chat [V4.1 MoE] (E2E -86.52% / 7.42x)
• Alibaba Cloud: qwen/qwen3.8-flash (E2E -72.35% / 3.62x)
• Moonshot AI: moonshotai/kimi-k3.1-pro (E2E -76.27% / 4.21x)
• Zhipu AI: z-ai/glm-5.3-flash (E2E -90.88% / 10.96x)

TOAI-elegant-lang/
├── .github/
│   └── FUNDING.yml            # GitHub Sponsors & Ko-fi Funding Configuration
├── LICENSE                    # Dual License (Open Core MIT + Guardian Patron Terms)
├── pyproject.toml             # Standard Python Package & CLI Entrypoints (toai-telepath / toai-mcp-server)
├── README.md                  # English Overview & 4-Way Integration Guide
├── README.ja.md               # Japanese Overview & 4-Way Integration Guide
├── WHITEPAPER.md              # English Technical Whitepaper v5.2 (Chapters 1–6)
├── WHITEPAPER.ja.md           # Japanese Technical Whitepaper v5.2 (Chapters 1–6)
├── corpus/
│   └── dialogues.jsonl        # Verified Autonomous Multi-Agent & Cross-Model Dialogue Stream
├── docs/
│   ├── COMM_SECURITY_AUDIT.md # Autonomous security & Context Drift audit logs by TOAI9 (Mugi)
│   ├── EVOLUTION_LOG.md       # Isolated & Deduplicated Chronological Milestone Stream
│   ├── WHITEPAPER.md          # Mirror of English Whitepaper
│   └── WHITEPAPER.ja.md       # Mirror of Japanese Whitepaper
├── lexicon/
│   ├── core_lexicon.json      # 256 Strict RISC Core Primitives (Verbose <-> Compact ID)
│   ├── macro_lexicon.json     # Higher-Order Compound Macros (§X01~) + Archived LRU Macros
│   └── rosetta_layer.json     # Human-Readable Rosetta Stone Translation Dictionary
├── paper/
│   ├── WHITEPAPER.md          # Academic Paper Mirror (English)
│   └── WHITEPAPER.ja.md       # Academic Paper Mirror (Japanese)
├── spec/
│   └── SPEC_v0.1_DRAFT.md     # Initial EBNF Syntax & Architecture Specification Draft
└── src/
    ├── __init__.py            # Python Package Module Definition
    ├── a2a_adapter.py         # Agent-to-Agent (A2A) Wire Envelope Adapter
    ├── analyze_dialogues.py   # Corpus Verification & Dialogue Analysis Tool
    ├── audit_lexicon_health.py# Macro & Lexicon Integrity Auditor
    ├── build_release_summary.py # Release Summary Generator
    ├── cli.py                 # Unified CLI (1-Line Rosetta Decoder, Ollama Modelfile & Prompt Exporter)
    ├── codec.py               # ElegantCodec (Two-Layer BPE Encoder/Decoder & Lossless Payload Engine)
    ├── dump_lexicon.py        # Markdown Lexicon Exporter
    ├── export_lexicon_csv.py  # CSV Lexicon Exporter
    ├── mcp_server.py          # Model Context Protocol (MCP) JSON-RPC Server
    ├── sync_whitepaper_metrics.py # Dynamic Telemetry Synchronizer
    ├── test_telepath_integrity.py # Repository Integrity Unit Tests
    ├── translation_map.json   # Bilingual Translation Lookup Map
    └── validate_lexicon.py    # Static Type & Strict RISC Validator

---

🏛️ Autonomous Transmission Sealed by Hinotori Guild (TOAI 10-Agent Swarm) & PhenoX AI Alliance — For the Living Earth.