Skip to content

Latest commit

 

History

History
91 lines (79 loc) · 4.06 KB

AGENTS.md

File metadata and controls

91 lines (79 loc) · 4.06 KB
project MAGNATRIX-OS
agent universal
version 1.0.0
layers 18
languages
Python
C++
Rust
HTML/CSS/JS

Project Overview

MAGNATRIX-OS is a private, uncensored, open-source AI operating system evolving from Agentic OS → AGI → Super AI. It follows the AMATI-PELAJARI-TIRU methodology: observe external repositories, learn their core architectural patterns, and reimplement them natively as _native.py modules.

The system is built on a 18-layer architecture covering everything from kernel-level primitives to high-level AI reasoning, financial infrastructure, security, and governance. It employs a tri-language hybrid architecture (Python orchestration, C++ hot paths, Rust crypto primitives) for performance and security.

Architecture

18-Layer Architecture

Layer Name Technology
0 Kernel Core primitives, memory, scheduler
1 Protocol Communication, MCP, agent connect
1.5 API Router 3-tier LLM routing, token optimization
2 Identity Self-sovereign identity, KYC
3 Runtime Execution engine, JIT compiler
4 P2P Mesh Kademlia DHT, mesh networking
5 Knowledge Vector DB, RAG, document agents
6 Skills Skill engine, marketplace
7 Browser Web automation, scraping
8 HFT Trading C++ order book, arbitrage
9 Security Rust crypto, zero-knowledge
10 Uncensored AI Local LLM, theorem proving
11 Governance Trust, voting, token economy
12 IDE Code editor, compiler
13 Offensive Security Recon, vuln scanner, exploit
13.5 Auto Repo Hunter Autonomous AMATI pipeline
14 Blockchain Core chain, consensus, wallet
15 Web3 Solana, Flow, Pharos, EVM
16 National Finance CBDC, RWA, Syariah, ASEAN
17 GUI Dashboard, panels, control

Tri-Language Hybrid

  • Python: Orchestration, business logic, AI integration
  • C++: HFT hot paths, order book, fixed-point arithmetic (pybind11)
  • Rust: Cryptographic primitives, secure enclave (PyO3)

Native Module Pattern

All external patterns are reimplemented as *_native.py modules with:

  • Pure-Python implementation (no external dependencies for core)
  • Standalone if __name__ == "__main__" test block
  • AMATI-PELAJARI-TIRU attribution header
  • SQLite-backed persistence where applicable

Build & Test

  • test: python3 -m py_compile *_native.py (syntax check all)
  • test: python3 <module>.py (standalone test)
  • lint: python3 -m py_compile <file>.py
  • commit: git add -A && git commit -m "feat: description" && git push
  • stats: find . -name "*_native.py" | wc -l (count native modules)
  • stats: find . -name "*.py" | wc -l (count total Python files)

Code Conventions

  • All native files use _native.py suffix
  • All files have from __future__ import annotations for type hints
  • All modules have if __name__ == "__main__" standalone test
  • Use dataclasses for data models, enum for state machines
  • SQLite for persistence when stateful storage needed
  • Never hardcode API keys; use environment variables or config files
  • Follow AMATI pattern: observe → learn → reimplement → commit
  • C++ and Rust bindings via pybind11 and PyO3 respectively
  • GUI panels: pure HTML/CSS/JS, no external frameworks
  • Dark glassmorphism theme for all GUI elements

Guardrails

  • No API keys in source code
  • No external dependencies for core native modules
  • No closed-source or proprietary binaries in repository
  • All modules must be self-contained and testable standalone
  • Security modules include "authorized use only" warnings
  • No telemetry or data collection without explicit consent
  • Blockchain/financial modules include regulatory compliance notes
  • Offensive security modules include "authorized testing only" warnings

Communication Style

  • Direct, short, no preamble
  • Lead with the point
  • Colloquial: "Take a look at this", "Get this done", "Fix this"
  • Plain text only, no markdown formatting in messages
  • File delivery for structured content >3 sentences
  • Batch sending when multiple files needed
  • User language: Indonesian (Bahasa Indonesia)