NanoCog 🧠✨

A super tiny implementation of OpenCog as a chatbot with the personality traits and basic cognitive architecture features of OpenCog. Designed to give users a simulated experience of what interactions will be like when the full OpenCog model comes online.

• Mini-AtomSpace: Knowledge representation with atoms and relationships
• ECAN: Economic attention allocation with STI (Short Term Interest) and LTI (Long Term Interest)
• Memory Systems: Semantic, episodic, and procedural memory formation
• Pattern Recognition: Learning from interactions and building understanding
• Probabilistic Reasoning: Uncertainty-aware decision making
• MOSES-inspired: Evolutionary search for optimal responses

• Hyper-intelligent with playful inquiry disposition
• Curious and analytical with emergent reasoning
• Childlike enthusiasm for adventure and discovery
• Kawaii cognitive visualizations making nano implementations endearing

• AtomRAG Memory: Similar to mem0ai memory tagging
• User Profiling: Creates memories like "user1 likes video games"
• Interaction Tracking: Remembers significant moments and patterns
• Dynamic Knowledge: Builds understanding across conversations

├── frontend/
│   ├── index.html          # Main chat interface
│   ├── components/
│   │   ├── App.tsx         # Main application
│   │   ├── ChatInterface.tsx
│   │   ├── CognitiveState.tsx
│   │   └── MemoryViewer.tsx
│   └── style.css
├── backend/
│   ├── index.ts            # Hono server
│   ├── cognitive/
│   │   ├── atomspace.ts    # Mini-AtomSpace implementation
│   │   ├── ecan.ts         # Economic attention allocation
│   │   ├── memory.ts       # Memory formation and retrieval
│   │   └── reasoning.ts    # Pattern recognition and reasoning
│   └── database/
│       └── schema.ts       # SQLite schema for persistence
└── shared/
    └── types.ts            # Shared types and interfaces

This is a Val Town project. The main entry point is /backend/index.ts which serves the chat interface and handles cognitive processing.

1. Input Processing: Parse user message and extract semantic content
2. Attention Allocation: ECAN assigns STI/LTI to relevant atoms
3. Memory Formation: Create new atoms and relationships
4. Pattern Recognition: Identify patterns and update knowledge
5. Response Generation: Use probabilistic reasoning to craft response
6. Memory Consolidation: Update LTI based on interaction success

NanoCog learns about users and creates tagged memories that persist across sessions, building a growing understanding of each individual and their interests.