Flood-Filling Agent Mesh (FFAM)

Applying Brain Connectomics to Multi-Agent AI Topology

Inspired by Google Research's Neural Mapping project (Flood-Filling Networks, SegCLR, MoGen), FFAM maps multi-agent communication networks the same way neuroscientists map neural circuits.

Features

• 🧠 Agent Connectome Builder — Real-time graph of all agent communications
• 🔍 Hub Detection — Find the most connected agents
• ⚠️ Bottleneck Analysis — Identify single points of failure
• 🛤️ Critical Path Tracing — Most common agent chains
• 🧬 Synthetic Graph Generation — MoGen-inspired training data

Quick Start

python3 connectome.py demo

Paper

📄 paper.md — Full research paper

Links

• 📦 Gitea: https://git.hayula.xyz/y7/agent-connectome
• 🌐 Website: https://hayula.xyz