Skip to content

Necream/NicCent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NicCent

State-driven open extension ecosystem powered by Coly.
AI should sleep — wake it only when needed.


🌐 Overview

NicCent is a reactive, state-driven extension ecosystem built on top of the Coly language and runtime.

It replaces traditional Agent / MCP / Workflow paradigms with a shared-state architecture based on ColyVar.

Instead of tool calls or linear pipelines, everything in NicCent is driven by state changes in a distributed network of extensions.


⚙️ Core Architecture

NicCent is composed of four fundamental components:

🧠 Coly (Language)

A programming language designed for state-driven computation.

👉 Official site: https://coly.spwaiter.top/


🏛 ColyServer (Runtime Core)

The central runtime responsible for:

  • Managing ColyVar (shared state)
  • Coordinating extensions
  • Broadcasting state changes
  • Maintaining system consistency

🔁 ColyVar (Shared State System)

ColyVar is the core abstraction of NicCent.

Everything is a state:

  • system metrics
  • AI outputs
  • extension results
  • scheduled events

Extensions react to changes in ColyVar instead of direct calls.


🤖 PAW (Powerful AI Widget)

PAW is the AI extension runtime of NicCent.

It provides:

  • AI-powered extensions
  • Multi-model support (via external adapters)
  • Reactive execution based on state changes

Supported models are NOT hardcoded — they are provided by extensions.

Examples:

  • OpenAI
  • DeepSeek
  • Claude
  • Qwen
  • Gemini
  • Local LLMs (Ollama, etc.)

⏱ Scheduler (Event System)

Scheduler is an independent process responsible for:

  • Time-based triggers
  • System event scheduling
  • Updating ColyVar states

No polling. No busy waiting.
Everything is event-driven.


🧩 Extension Ecosystem

NicCent is built as a network of extensions, not a hierarchy.

Example extensions:

  • PAW (AI runtime)
  • DiskGuard (storage monitoring)
  • MemoryOcean (memory analytics)
  • ThermalSense (hardware state tracking)
  • Scheduler (event engine)

Extensions communicate through ColyVar state changes, not direct calls.


🧠 Key Concept: State-Driven Architecture

Instead of:

  • tool calls
  • agent chains
  • workflows

NicCent uses:


State Change → Reaction → State Update → Network Propagation

Everything is reactive.


🚫 Not Like Other Systems

NicCent is NOT:

  • ❌ An agent framework
  • ❌ An MCP implementation
  • ❌ A workflow automation system

It is a reactive distributed runtime ecosystem.

Paradigm Model
MCP Tool calling
Agent Continuous reasoning loop
Workflow Sequential execution
NicCent State-driven reactive network

🌌 System Topology

NicCent is not a tree — it is a network graph.


PAW

↗ ↑ ↖


DiskGuard ColyServer MemoryOcean
↘   ↓   ↙
ThermalSense
↓
Scheduler

All nodes interact via shared state (ColyVar).


⚡ Key Principles

  • AI should sleep when idle
  • No unnecessary computation
  • No centralized control flow
  • Everything is reactive
  • Everything is state

🔓 Open Source Model

  • NicCent Core: Open Source
  • PAW: Open Source
  • Extensions: Open or Closed (developer choice)

🛣 Roadmap

  • MCP Adapter Layer
  • Extension Marketplace
  • Distributed ColyServer
  • Cross-platform runtime support
  • Cloud synchronization layer (optional)

🚀 Getting Started

(to be defined in implementation phase)

Planned:

  • install Coly runtime
  • run ColyServer
  • register PAW extension
  • start Scheduler process

💡 Philosophy

"AI is not something that runs continuously.
It is something that should sleep until the world changes."

NicCent is built around this idea.


📎 Related


📄 License

TBD (Recommended: Apache-2.0 or MIT)

About

Build extensions, not agents. NicCent is a state-driven runtime ecosystem powered by Coly, where AI (PAW), schedulers, and system extensions react through shared state instead of requests.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages