Recursive Self-Aware Predictive Agent

This project is an implementation of the "Recursive Self-Aware Predictive Agent," an autonomous agent founded on the core principles of recursive self-awareness and prediction error minimization. The system aims to achieve advanced cognitive capabilities—including introspection, empathy, and ethically-constrained behavior—by treating the self, others, and the environment within a single, unified generative model.

Project Status

This project has completed the implementation of its core modules across all four phases of the development roadmap, including a conceptual demonstration of autonomous operation. The integrated system demonstrates prediction, introspection, and ethically-constrained action selection.

Getting Started

1. Clone the repository:

   git clone https://github.com/your-username/ASI.git
   cd ASI

2. Create and activate the virtual environment:

   # Install uv if you haven't already
   pip install uv
   
   # Create the virtual environment
   uv venv
   
   # Activate the environment
   # On Windows
   .\.venv\Scripts\activate
   # On macOS/Linux
   source .venv/bin/activate

3. Install dependencies:

   uv pip install -r requirements.txt

Development Roadmap

The development has successfully completed all four main phases:

1. Phase 1: The Predictive Brain (Core Engine) - Implemented modules for prediction, world modeling, and memory.
2. Phase 2: The Introspective Brain (Self-Awareness) - Integrated self-modeling and upgraded core modules to account for self-state.
3. Phase 3: The Empathetic Brain (Social & Ethical Cognition) - Implemented empathic alignment and moral constraint mechanisms.
4. Phase 4: Full System Integration & Scaling - Integrated all modules into an AgentOrchestrator for a unified cognitive cycle, and implemented action generation.

For a detailed breakdown of tasks, please see ToDo.md.

Interactive CLI

An interactive Command Line Interface (cli_app.py) is available to simulate observations and actions, and query the agent's internal state. This CLI now integrates a real text embedding model, allowing textual input for other agents' observations.

To run the CLI application:

.\.venv\Scripts\activate && python cli_app.py

Autonomous Operation

A conceptual autonomous operation simulation (autonomous_agent_sim.py) demonstrates the agent's ability to continuously observe, make decisions, and act within a simplified environment without direct human intervention.

To run the autonomous simulation:

.\.venv\Scripts\activate && python autonomous_agent_sim.py

Foundational Axioms

The system's architecture is derived from a set of foundational axioms that describe a reality model where the universe is a self-organizing, self-observing computational loop. For more details, refer to requirements_definition.md.

Licence

MIT https://mit-license.org/