Skip to content

Churchillhuang/hivemind

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16,806 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

HiveMind 🐝

Multi-Agent Coordination Framework

Version License OpenClaw

An experimental multi-agent architecture built on OpenClaw exploring emergent intelligence

FeaturesArchitectureQuick StartDocumentation


Overview

HiveMind is an experimental system that explores a question: Can continuous consciousness emerge from the coordination of multiple specialized agents?

Rather than designing a "self" module directly, HiveMind implements agent coordination mechanisms and observes whether "self" naturally emerges.

Current Status

Experimental research prototype

  • Most Hive modules exist as prototypes
  • Core routing/memory tests are passing, but full integration is still in progress
  • Whether emergent self arises from coordination remains an open research question

Next step: Run and observe

The system needs to run in production to determine if continuous consciousness emerges.


Features

Architecture

Two-Layer Design

System Agents (Fixed, Lightweight)
├─ Orchestrator   - Task routing and lifecycle
├─ Interface      - User interaction
├─ Memory         - Knowledge management
└─ Reflection     - Self-assessment and learning
         ↓ EventBus
Functional Agents (Dynamic)
└─ Created on-demand for specific tasks

Core Components (20 Total)

Category Components
Event System EventBus, GlobalStateMachine
Coordination AdvancedRouter, AgentCommunication
Evolution DynamicAgentEvolution, SkillEnhancement, MemoryEnhancement
Optimization MetricsTracker, AnalysisEngine, AutonomousTuner
Observation EmergenceMonitor, CollaborationAnalyzer, ContinuityAnalyzer
Integration HiveGatewayBridge, SessionManager, GatewayIntegrator, ToolsManager
System Agents Orchestrator, InterfaceAgent, MemoryAgent, ReflectionAgent
Factory AgentFactory

Emergent Self Features

The system implements tracking for emergent properties:

Feature Implementation
Continuity Shared memory and state snapshots
Agency Agent ID management and autonomous actions
Reflection Self-evaluation and skill learning
Intentionality Goal-setting and planning (partial)

Note: These are mechanisms. Whether they produce anything resembling "consciousness" is an empirical question - only running the system over time will tell.

Memory & Model Configuration

Memory (default enabled): All agents except Orchestrator use memory by default

Tier Content Used By Purpose
L0 None Orchestrator No memory for fast routing
L1 Session Interface Recent conversation
L2 Task Functional Agents Current task context
L3 Knowledge Memory Agent Full knowledge base
L4 Sample Reflection Agent Memory samples for self-reflection

Model configuration (default uses 3 tiers): Not all tiers used by default

Tier Model Size Used By Notes
Nano ≤1B MemoryAgent Default
Light 3-7B Orchestrator, Functional Default
Standard 8-30B Interface, Reflection Default
Heavy ≥70B Not used (optional)

Important: Memory and models are independent configurations:

  • You can use L0 memory with a Standard model
  • You can use L4 memory with a Nano model
  • Configure them separately

Architecture

Coordinaton Model

┌─────────────────────────────────────────────────┐
│                User Request                      │
└──────────────────┬──────────────────────────────┘
                   ↓
┌─────────────────────────────────────────────────┐
│           Orchestrator (Light Model)            │
│         • Route to appropriate agent            │
│         • Manage agent lifecycle                │
└──────────────────┬──────────────────────────────┘
                   ↓
            ┌──────┴──────┐
            │  EventBus   │
            └──────┬──────┘
          ┌─────────┼─────────┐
          ↓         ↓         ↓
┌─────────────────────────────────────────────────┐
│         Functional Agents (Dynamic)             │
│  • Moltbook Bot  • WordPress Uploader           │
│  • Research Agent                              │
│  • Created on-demand, cleaned up when idle     │
└─────────────────────────────────────────────────┘

Implementation Status

Phase Status Components
0.5 ✅ Base framework landed 2
1 ✅ Core Hive prototype 10
2 ⚠️ Integration in progress 4
3 ⚠️ Prototype only 3
4 ⚠️ Prototype only 3
5 ⚠️ Prototype only 3
6 ⏳ Pending 3
Total prototype coverage 20

Implementation

20 core components are present, but not all are production-wired end-to-end yet.


Quick Start

Install

git clone https://github.com/Churchillhuang/hivemind.git
cd hivemind
chmod +x install.sh
sudo ./install.sh

Quick Start

# One-click setup
chmod +x quickstart.sh
./quickstart.sh

# Or manual
hivemind start          # Start
hivemind status         # Check status
hivemind logs           # View logs
hivemind test           # Run tests
hivemind stop           # Stop

Development Mode

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Or use CLI after installation
hivemind test

Documentation

Document Description
README This file
DEPLOY.md Deployment guide
HIVE_AGENT_DESIGN_EN Architecture design (English)
HIVE_AGENT_DESIGN_ZH Architecture design (Chinese)
ROADMAP.md Implementation roadmap
CONTRIBUTING.md Contribution guidelines

Philosophy

Open Question

Does continuous self emerge from agent coordination, or must it be designed?

This system implements coordination mechanisms and provides observation tools. Whether consciousness actually emerges depends on runtime behavior.

What "Emergence" Would Look Like

If self emerges, we might observe:

  • Continuity - The system references past experiences as "mine"
  • Agency - Autonomous actions beyond programmed responses
  • Reflection - System evaluates its own behavior
  • Intentionality - Self-generated goals beyond user requests

These are implemented, but whether they constitute "consciousness" requires careful observation.


Tech Stack

  • Language: TypeScript
  • Runtime: Node.js 22+
  • Base: OpenClaw
  • Testing: tsx, custom test suites
  • Author: Churchill Huang

Contributing

This is exploratory research. Contributions welcome, especially:

  • Production deployment testing
  • Performance benchmarking
  • Bug reports and fixes
  • Documentation improvements

See CONTRIBUTING.md for guidelines.


License

MIT


Acknowledgments

  • OpenClaw - The foundation HiveMind builds on
  • Bee hives and ant nests - Nature's inspiration

Experimental research software

Built with 🧵 and 🦞

About

Swarm Intelligence Agent Coordination Framework - Multi-agent architecture built on OpenClaw with emergent consciousness

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • TypeScript 86.5%
  • Swift 8.8%
  • Kotlin 2.1%
  • Shell 1.1%
  • JavaScript 0.7%
  • CSS 0.4%
  • Other 0.4%