Cnexus Cos Core V2

SKILL.md

---
name: cnexus-cos-core-v2
description: >-
  CNexus 2.0 — Personal Pure Edition.
  Holographic Memory · 6-Step Cognitive Loop · Local-First Brain-Like Architecture.
  Install this skill to run the complete L0 cognitive kernel locally:
  Observe → Cognize → Decide → Speak → Store → Reflect.
---

# 🧠 CNexus 2.0 - Personal Pure Edition

> **Holographic Memory · 6-Step Cognitive Loop · Local-First Brain-Like Architecture**

---

## Project Introduction

CNexus 2.0 is a lightweight, fully runnable, **local-first personal cognitive operating system prototype**. It focuses on achieving a complete single-machine closed-loop cognitive process: **Observe → Cognize → Decide → Speak → Store → Reflect**.

Instead of relying on cloud LLMs or complex distributed systems, CNexus 2.0 prioritizes running a stable and soulful **L0 Kernel** on a personal computer. The current Personal Pure Edition has successfully implemented:

- **6 Pure-Function Cognitive Kernel** (`observe_reducer` → `cognize_reducer` → `decide_reducer` → `speak_reducer` → `store_reducer` → `reflect_reducer`)
- **BlockStore Holographic Memory System**: Supports multiple memory block types including `emotion` (single-slot overwrite), `episodic` (append-only), `semantic` (REM-compressed), `code_class`, `code_function`, `vision_component`, etc.
- **Dynamic Memory Nebula Visualization**: Canvas 2D + force-directed layout with **wormhole** semantic links, activation diffusion, and crystallization layout — your memory literally grows, connects, and breathes on screen.
- **REM Deep Sleep Mechanism**: Periodic pruning, compression, and semantic consolidation of long-term memories into high-importance `sem-rem-*` blocks.
- **Code & Vision Projection**: Automatically projects project code structure and image inputs into the memory graph, enabling the system to "understand its own codebase."
- **Strong Local-First Design**: Defaults to local Ollama models (`nomic-embed-text` + Llava-class vision). Cloud APIs are only used when explicitly configured via environment variables.

## Key Highlights

- **Personality Stability & Sovereignty**: Fixed identity anchors, single-slot emotion/goal, protected long-term facts, and anti-drift mechanisms.
- **Living Memory Topology**: Real-time activation decay, wormhole resonance, and semantic diffusion.
- **Full Traceability**: Every cognitive cycle is recorded as structured traces with causal replay support.
- **Minimalist Deployment**: Single-process runtime (`app_v2.py` + frontend), no Docker, no database, no WebSocket required.

## Current Status

The **Personal Pure Edition** has achieved a complete L0 single-machine cognitive closed loop, including activation diffusion, wormholes, REM consolidation, and multimodal/code projection. It is ready for local deployment, experimentation, and deep hacking.

Enterprise/distributed capabilities (DomainStorageAdapter, CRDT, Spine dynamic execution graphs, cluster consensus, etc.) currently exist as Shadow projections and are not yet wired into the main memory topology. Community contributions toward decentralization are highly welcome.

## Who It's For

- Developers and researchers interested in **embodied cognition**, **memory systems**, **self-evolving AI**, and **long-term personality stability**
- Anyone who wants a personal AI companion that truly **thinks, remembers, reflects, and visualizes its own mind**
- Cognitive architecture explorers building the next generation of sovereign, local AI systems

---

**Project Positioning**: Not another ChatGPT wrapper — but a **fully controllable personal cognitive kernel laboratory**.

## 📦 Installation

From ClawHub:
```bash
clawhub install cnexus-cos-core-v2
```

## 🚀 Quick Start

```bash
# 1. Start CNexus runtime
python app_v2.py

# 2. Open in browser
open http://127.0.0.1:7865

# 3. Start conversing — watch your memory nebula grow
```

## 📡 API Endpoints

| Endpoint | Description |
|---|---|
| `POST /api/converse` | Cognitive loop — distill interaction |
| `GET /api/status` | System state + memory graph |
| `GET /v1/kernel/record/{trace_id}` | Full execution trace replay |

## 🔗 Links

- **GitHub:** [https://github.com/plusunm/CNexus2.0](https://github.com/plusunm/CNexus2.0)
- **Runtime:** `http://127.0.0.1:7865` (default)
- **Tech Stack:** Python + TypeScript/React + Canvas 2D + Local LLMs (Ollama)