@sym-bot/core

Cognitive core for SYM mesh nodes — SVAF fusion, CMB encoding, xMesh intelligence.

This is the internal engine used by SYM. It provides:

• CMB Encoder — creates Cognitive Memory Blocks with CAT7 fields, content-addressable keys
• SVAF Evaluator — per-field evaluation of incoming signals (neural + heuristic)
• xMesh — per-agent Liquid Neural Network (CfC) with bimodal τ dynamics
• Context Encoder — encodes text into hidden state vectors (h1, h2)
• Coupling Engine — drift computation, per-neuron blending, Kuramoto metrics

Usage

Used as a dependency of SYM. Not typically installed directly.

# Clone both repos side by side
git clone https://github.com/sym-bot/sym.git
git clone https://github.com/sym-bot/sym-core.git
cd sym && npm install

Trained Models

The models/ directory contains pre-trained models:

• svaf_v2.pt — SVAF per-field evaluation (78.7% accuracy)
• xmesh_v4.pt — per-agent LNN (AUC 0.83, bimodal τ: 64 fast / 64 slow)

These are proprietary trained weights. The training code and data are not included.

Specification

This implementation follows MMP v0.2.0.

License

Apache 2.0 — see LICENSE

SYM.BOT Ltd