TradeFlow Command System

AI-Powered Multi-Agent Trade Intelligence for African Importers and Exporters

Built for the Microsoft Agents League Hackathon 2026 — Reasoning Agents Track

---

What is TradeFlow?

TradeFlow is a multi-agent AI system that helps African SME importers and exporters make high-confidence cross-border trade decisions.

It does not give you a number. It gives you a decision — with reasoning, confidence levels, risk flags, and a full audit trail of every assumption made.

Example query:

"I want to import 500 units of Samsung smartphones from China to Lagos via Apapa port."

TradeFlow responds with:

• Market intelligence analysis (FX risk, duties, demand signals)
• Logistics routing options (Apapa vs Tincan vs Cotonou)
• Compliance and risk assessment (SONCAP, NCC, customs requirements)
• A verified final trade recommendation (PROCEED / PROCEED WITH CAUTION / PAUSE)
• A full decision traceability report showing every assumption and data flag

---

The Crisis Commander Loop

TradeFlow is not a static query system. It is a dynamic re-planning intelligence system.

When a disruption signal arrives — a port closure, FX spike, policy change, or supplier fraud alert — TradeFlow automatically re-runs the entire agent chain with the new context and produces a revised recommendation.

---

Agent Architecture

• Orchestrator Agent: decomposes query, routes to specialists, manages state
• Market Intelligence Agent: FX rates, duties, demand signals, landed cost
• Route and Logistics Agent: port conditions, freight options, transit times
• Risk and Compliance Agent: SONCAP, NCC, NAFDAC, supplier risk, FX risk
• Critic and Verifier Agent: cross-checks all outputs, scores confidence, final verdict

---

Responsible AI Design

• Every data point flagged as ESTIMATED or ASSUMED
• Every assumption listed explicitly
• Confidence levels assigned per agent
• No fabricated tariff rates or regulatory requirements
• Traders directed to verify with SON, NAFDAC, NCC, Nigeria Customs

---

Tech Stack

• Platform: Microsoft Azure AI Foundry
• Model: gpt-4.1-mini (East US 2)
• SDK: azure-ai-projects, openai Python SDK
• Authentication: Azure DefaultAzureCredential
• Language: Python 3.11+

---

Setup

export AZURE_OPENAI_ENDPOINT="your_azure_openai_endpoint"
export MODEL_DEPLOYMENT_NAME="gpt-4.1-mini"
pip install openai azure-ai-projects azure-identity
python3 tradeflow.py

---

Hackathon Track

Track: Reasoning Agents — Microsoft Foundry Additional category: Hack for Good

---

Built by

Mary Ezinne Obasi (Ria-Zee) AI Automation and Agentic Engineer Lagos/Abuja, Nigeria