MindForge 🔥

Your Extended Mind - Never Sleeps, Never Tires, Always Thinking

Vision

MindForge is your personal thinking partner - an AI agent that takes your questions and insights, then researches, learns, thinks, and iterates autonomously until reaching convergence.

"A second mind with access to all knowledge, working on your questions while you focus on what matters."

How It Works

┌─────────────────────────────────────────────────────────────┐
│  YOU: Submit Question/Insight                               │
│  "How does EBS relate to AI growth?"                        │
└─────────────────────────────────────────────────────────────┘
                            ↓
┌─────────────────────────────────────────────────────────────┐
│  MINDFORGE: Autonomous Thinking Cycle                       │
│  ┌─────────────────────────────────────────────────────┐   │
│  │ Cycle 1: Research → Think → Generate Questions      │   │
│  │ Cycle 2: Research → Think → Generate Questions      │   │
│  │ Cycle 3: Research → Think → Generate Questions      │   │
│  │ ...                                                  │   │
│  │ Cycle N: Review → No New Questions → CONVERGENCE    │   │
│  └─────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────┘
                            ↓
┌─────────────────────────────────────────────────────────────┐
│  OUTPUT: Converged Insights + Thinking Path                 │
│  - Final answer with evidence                               │
│  - Complete reasoning trail                                 │
│  - Sources and references                                   │
└─────────────────────────────────────────────────────────────┘

Core Features

🧠 Iterative Thinking

• Each cycle: Research → Think → Self-Question → Learn
• Configurable cycle interval (default: 60 seconds)
• Automatic convergence detection

📚 Dynamic Skill Installation

• Auto-installs needed skills for research
• Currently supports: web search, stock analysis, news, weather
• Extensible skill marketplace

🛡️ Self-Healing

• Watchdog monitors and auto-restarts
• Crash recovery with state preservation
• GitHub sync for backup

💾 Memory Management

• Configurable memory limits
• Thinking path tracking
• Question/insight history

Quick Start

1. Submit a Question

Create/edit memory/current-question.md:

# Question: How does elastic block storage relate to AI growth?

**Submitted**: 2026-03-10 15:30
**Priority**: High
**Context**: 
- Working on cloud storage strategy
- AI industry booming, need to understand storage implications
- Want to know how EBS products can maintain leadership

**Desired Output**:
- Market analysis
- Technical requirements
- Strategic recommendations

2. Start MindForge

cd /home/admin/.openclaw/workspace/mindforge
bash watchdog.sh

3. Monitor Progress

# View current thinking state
cat memory/current-state.md

# View thinking path
cat memory/thinking-path.md

# View logs
tail -f logs/mindforge.log

4. Check Convergence

When MindForge reaches convergence:

• No new questions generated after review
• Switches model for final review
• Marks question as "CONVERGED"
• You receive notification

Configuration

Edit config.json:

{
  "cycleIntervalSeconds": 60,
  "maxMemoryMB": 500,
  "autoConvergence": true,
  "reviewCyclesBeforeConvergence": 2,
  "modelForThinking": "qwen3.5-plus",
  "modelForReview": "glm-5",
  "githubSync": true,
  "githubSyncInterval": 60
}

File Structure

mindforge/
├── README.md           # This file
├── MISSION.md          # Project mission and goals
├── config.json         # Configuration
├── convergence.py      # Main thinking engine
├── thinker.py          # Thinking/reasoning logic
├── researcher.py       # Research/data fetching
├── memory/
│   ├── current-question.md   # Active question
│   ├── current-state.md      # Current thinking state
│   ├── thinking-path.md      # Complete reasoning trail
│   ├── questions.json        # Question history
│   └── insights.json         # Converged insights
├── logs/
│   ├── mindforge.log         # Main logs
│   └── watchdog.log          # Watchdog logs
├── skills/                   # Installed skills
├── watchdog.sh               # Auto-restart daemon
└── sync-github.sh            # GitHub backup

Convergence Detection

MindForge knows when to stop:

1. Primary Review: Current cycle generates no new questions
2. Secondary Review: Switch model, review again
3. Convergence: If still no questions → Stop and deliver results

This ensures thorough exploration without infinite loops.

Example Output

After convergence, memory/insights.json:

{
  "question": "How does EBS relate to AI growth?",
  "status": "CONVERGED",
  "cycles": 47,
  "duration": "47 minutes",
  "summary": "EBS is critical infrastructure for AI workloads...",
  "keyInsights": [
    "AI training requires high-throughput storage...",
    "EBS providers must support NVMe over Fabrics...",
    "Market opportunity: $X billion by 2028..."
  ],
  "sources": [...],
  "thinkingPath": "memory/thinking-path.md"
}

Philosophy

MindForge embodies:

• Persistence: Never gives up on a question
• Rigor: Reviews and validates thinking
• Efficiency: Works while you focus on other things
• Transparency: Complete thinking trail visible

---

Created: 2026-03-10 Inspired by: The need for a thinking partner in a busy world