DEV Community

Damjan Žakelj
Damjan Žakelj

Posted on

Visualizing Trust in Multi-Agent Systems — The Swarm-ISM-X Public Demo (v2)

For the past months I’ve been experimenting with ways to visualize trust and stability in distributed AI systems — the kind of architectures where dozens of agents must cooperate without a central brain.

The result is something I call Swarm-ISM-X.

The Public Demo (v2) is now open-sourced — a clean, safe version that shows how the swarm behaves, not why it behaves that way.

🌀 What you’ll see

A Tkinter-based GUI that displays 10 agents along a horizontal line.

Each agent moves, stabilizes, and maintains formation under light “wind” disturbances.

Each agent has a “passport” indicator (green = valid, red = invalid).

An “Auto Demo” mode runs scripted sequences for presentations.

The simulation updates in real time — you can watch the system find balance, lose it, and regain it.

🔍 What’s really happening

Under the hood, each agent is governed by a simplified consensus-like controller:

Controller

$$
u_i \;=\; -\,k_i \nabla_i S
$$

where $S$ is a constraint vector maintaining equal spacing and total span.

The real ISM-X framework extends this idea with:

Adaptive gain tuning using resonant feedback (not in public demo).

Cryptographic attestation (Ed25519 + HMAC commitments).

Passport issuance and verification between agents.

Log-periodic modulation for stability over communication delays.

The public demo keeps only the first-order visible dynamics — enough to show formation control and disturbance recovery — while replacing sensitive parts with lightweight placeholders.

🔒 What’s included vs. hidden
Layer Included Hidden
GUI visualization ✅ –
Swarm dynamics (simple consensus) ✅ –
Passport system (stubbed SHA-1) ✅ Real attestation (Ed25519/HMAC)
Adaptive control & resonance ❌ proprietary
Informational geometry layer ❌ research
⚙️ Run it yourself
git clone https://github.com/Freeky7819/swarm-ismx-gui-demo.git
cd swarm-ismx-gui-demo
pip install numpy
python main_gui_public.py

Works out of the box on Python 3.10+.
The GUI shows live values of (||S||), (J), and per-agent gains (k_i).

🧩 Why it matters

Visual demos like this help bridge AI orchestration and trust architectures.
You can see — literally — what happens when an agent’s integrity fails, when noise enters, or when collective damping stabilizes the system.

This isn’t a neural network or RL — it’s a physically grounded, interpretable control system.
Think of it as a way to watch trust itself breathe.

GitHub: Swarm-ISM-X GUI Demo v2

Author: Damjan
Reason in resonance.

Feedback is always welcome — especially if you work on:

multi-agent coordination,

real-time visualization,

control theory + cryptographic verification bridges.

Let’s make AI agents not only smarter — but also more honest.

Top comments (0)