metisos-arc-core 1.2.0

Adaptive Recursive Consciousness (ARC) Core - A framework for continual learning AI systems

ARC Core

Adaptive Recursive Consciousness (ARC) Core is a framework for building continual learning AI systems that can learn and reason over time. It features:

• Continual Learning: Real-time learning with LoRA adapters
• Reasoning Engine: Graph-based reasoning and pattern recognition
• Biological Learning: Implements biological learning mechanisms
• Model Agnostic: Works with various transformer architectures

Installation

pip install arc-core

Quick Start

from arc_core import LearningARCConsciousness, ARCCore, ARCTrainer  # All aliases for the same class

# Initialize with a base model
arc = LearningARCConsciousness(model_name="gpt2")

# Learn from interactions
arc.learn_from_experience("The sky appears blue due to Rayleigh scattering")

# Generate responses
response = arc.generate("Why is the sky blue?")
print(response)

Features

• Dynamic LoRA Adapters: Automatically adapts to different model architectures
• Reasoning Graph: Maintains a knowledge graph of learned concepts
• Biological Learning: Implements contextual gating and cognitive inhibition
• Persistence: Saves learning progress between sessions

Documentation

For detailed documentation, see ARC Core Documentation.

License

MIT