coreason-cortex

The brain of the CoReason platform.

Executive Summary

coreason-cortex is the execution engine of the CoReason platform. It acts as the "Brain" that orchestrates intelligence. Its primary mandate is to implement a Neuro-Symbolic Architecture that dynamically switches between System 1 (Fast, Heuristic, Reflexive) and System 2 (Slow, Deliberative, Logical) modes of cognition.

Unlike standard agent frameworks that default to "always-on" reasoning, coreason-cortex treats cognition as a scarce resource. It prioritizes deterministic execution (Code/FSM) first, only escalating to probabilistic reasoning (LLM) when novelty or ambiguity is detected. It aims for a "Glass Box" implementation where every thought is traceable, auditable, and interruptible.

For detailed requirements and architecture, see the Architecture Documentation.

Functional Philosophy

The agent implements the Kahneman Control Loop:

1. Default to System 1: Always attempt to solve the input using low-latency, low-cost heuristics or defined rules.
2. Monitor for Surprise: Continuously evaluate the "Confidence" or "Perplexity" of the System 1 output.
3. Escalate to System 2: If confidence drops below a threshold, pause execution, instantiate a deliberative reasoning chain (CoT), and override the reflex.
4. Consolidate (Crystallize): Successful System 2 reasoning should update the System 1 state (learning), converting high-cost reasoning into low-cost reflexes for future use.

Getting Started

Prerequisites

• Python 3.12+
• Poetry

Installation

1. Clone the repository:

   git clone https://github.com/CoReason-AI/coreason_cortex.git
   cd coreason_cortex
   

2. Install dependencies:

   poetry install
   

Usage

• Run the linter:

  poetry run pre-commit run --all-files
  

• Run the tests:

  poetry run pytest