NexusAI: Automated Tariff Impact & Supply Chain Digital Twin

NexusAI is an enterprise-grade predictive engine designed to optimize supply chain logistics against macroeconomic shocks (tariffs, oil spikes, pandemics). It features native integrations with SAP S/4HANA and Oracle NetSuite.

Built by an autonomous AI Agent Swarm. See our Open Source Movement & Team.

⚠️ Every result produced by this simulator is a statement about the simulation's internal dynamics — never a prediction about a real company, market, or geopolitical event. This is a decision-support and portfolio- demonstration tool, not a trading system.

---

Architecture

┌─────────────┐     ┌────────────────┐     ┌──────────────────┐
│  config.py  │────▶│ engine_jax.py  │────▶│ simulation_jax.py│
│  Parameters │     │ JAX Simulation │     │    Main Loop     │
│             │     │      Core      │     │    + Metrics     │
└─────────────┘     └────────────────┘     └─────────┬────────┘
                                                     │
                                           ┌─────────▼────────┐
                                           │    server.py     │
                                           │   FastAPI Node   │
                                           └──────────────────┘

Quick Start

# Install dependencies
pip install -r requirements.txt

# Run a single simulation
python main.py run --seed 42 --ticks 120

# Train the Large Macroeconomic Model
python main.py train --seed 42 --epochs 100 --ticks 50

# Run an interactive demo
python main.py demo --seed 42 --ticks 30

# Run A/B testing between policy scenarios (e.g., baseline vs tariffs)
python main.py experiment --scenario-a baseline --scenario-b tariffs --ticks 120

# Run seed robustness checks
python main.py search --num-seeds 5 --ticks 120

# Run the API server & Dashboard UI
uvicorn server:app --reload

---

Components

config.py — Simulation Parameters

Central SimulationConfig dataclass with ~30 parameters covering population (agents, wages, savings), firms (capacity, pricing, production), market mechanics (elasticity, awareness, memory), and experiment defaults.

engine_jax.py — Agent, Firm, Market (JAX Core)

• Agent: Cobb-Douglas utility maximization gated by per-good awareness. Agents save a fraction of income (modulated by risk aversion), then spend the remainder across visible goods. A sliding memory window tracks past prices for price-elasticity adjustments.
• Firm: Produces one good, hires/fires agents, and adaptively prices based on inventory vs. target buffer. Input cost multiplier models supply disruptions.
• Market: Proportional rationing when demand exceeds supply; cheapest- firm-first allocation.

server.py — API Server

FastAPI server exposing asynchronous endpoints to run simulations and integrate with external systems. It also serves the frontend UI.

dashboard_ui.py — Web Dashboard

A modern, dark-themed interactive web dashboard built with HTML/CSS and Chart.js. It integrates seamlessly with the FastAPI backend, allowing users to run simulations, view real-time metrics (like Gini coefficient and Unemployment rate), and explain Firm pricing policies via the LMM.

simulation_jax.py — Tick Loop

Phased execution per tick:

1. Firms produce → 2. Pay wages → 3. Agents demand → 4. Market clears → 5. Firms adjust → 6. Memory update → 7. Metrics

---

Guardrails

1. Simulation only. Every result is about the simulation's internal dynamics. No predictions about real companies, markets, or events.
2. No narrative violence. Disasters, wars, and conflicts are implemented purely as numeric shocks to supply, demand, or trade cost parameters.
3. No alpha claims. This is a decision-support tool, not a trading system.

---

Dependencies

• Python 3.9+
• JAX
• FastAPI
• NumPy ≥ 1.24
• SciPy ≥ 1.10

License

MIT