A local-first virtual intelligence system designed to perceive, remember, reason, and act within an ethical framework.
VICy is designed as a local, transparent alternative to opaque AI systems. It focuses on structured memory, ethical behavior, and real-world integration.
- Modular architecture (Core, Memory, Sense, Act, Ethos)
- Associative memory graph (transparent knowledge representation)
- Local-first design (privacy + efficiency)
- Ethical guard system (VicyEthos)
- Persistent context across sessions (VicyContinuity)
VICy (Virtual Intelligence) is a local-first open cognitive system designed to perceive, remember, and act within a structured and ethical framework. It combines a modular architecture with an associative memory graph, enabling transparent reasoning, context awareness, and adaptive behavior.
Unlike traditional AI systems, VICy focuses on clarity, control, and real-world integration. It is designed to operate efficiently on local hardware, respect user privacy, and support humans and other life forms through responsible and verifiable decision-making.
The project follows a long-term vision of building a system that connects perception (VICySense), memory (VICyMemoryGraph), reasoning (VICyCore), ethics (VICyEthos), and action (VICyyAct) into a coherent and extensible platform.
Early development (v0.1.x.x)
Currently working on:
- Base libraries (Network Layer, Terminal Layer, etc.) -> IN PROGRESS
- MemoryGraph basics
- Query system
- Persistence -> IN PROGRESS
| Target/Configuration | Status |
|---|---|
| Linux/Default |  |
- Virtual Intelligence instead of Artificial Intelligence
- Transparency over black-box behavior
- Local execution (privacy + efficiency)
- Ethical constraints as core system layer
- Long-term, modular system design
- Inspired by practical engineering (Donald E. Knuth, C inventors, hacker culture, etc.)
- Decision-making
- Query handling
- Orchestration
- Concept nodes
- Relations
- Long-term memory (disk)
- Short-term memory (RAM)
- Guard system
- Rule validation
- Value-based behavior control
- Context persistence across sleep/shutdown
- Snapshot of active state
- Sensor input (temperature, pressure, etc.)
- Signal -> perception mapping
- Interacting with the world
- Actors (relays, motors, etc.)
- Active context
- Recent interactions
- Current goals
- Temporary associations
- Concepts
- Definitions
- Learned relations
- Persistent knowledge
- Content-based hashing (SHA-256)
- Nodes + relations stored on disk
- Indexes for fast lookup
- ID (hash)
- Label
- Type
- Confidence
- Metadata
- From -> to
- Type (is_a, related_to, etc.)
- Weight/strength
- What is X?
- Define X
- What is related to X?
- How is X connected to Y?
- Query handling
- Text-based protocol (initial)
- status
- suspend
- sleep
- resume
- save memory
- logs
- RUNNING
- IDLE
- SUSPENDING
- SLEEPING
- RESTORING
Before sleep:
- Save relevant context (not full RAM)
After wake:
- Restore active context
- Resume reasoning
- Basic MemoryGraph
- Node and relation storage
- Simple CLI queries
- Disk persistence
- Context memory
- Improved query engine
- Control interface
- Sensor integration (VICySense)
- Acting with the environment (VICyAct)
- Internal state model
- Learning system
- Optimization (C/Assembly/FPGA)
- Advanced reasoning
Core system:
- GPLv3
Optional modules/interfaces
- Permissive (ISC/MIT)
- GNU make
- C++ compiler with at least C++20 support
$ git clone git@github.com:Krotti83/VICy.git
$ cd VICy/src/VICy
$ make
- Keep system modular
- Avoid over-engineering early
- Build -> test -> refine
- Optimize later
This project was developed with the assistance of ChatGPT, used as a tool for idea refinement, architectural design, and discussion.