An Agentic AI Supply Chain Hedge Fund for SMEs
TradeJack levels the playing field for local vendors by autonomously predicting market shocks, detecting monopoly price-fixing, and executing split-second procurement across legacy B2B portals before prices spike.
Small and Medium Enterprises (SMEs) lack the custom APIs and massive procurement teams that Fortune 500 companies possess. When global supply chain shocks occur (e.g., floods, port strikes), local vendors find out too late, losing 20-30% of their margins to artificially inflated wholesale prices and hidden monopolies.
TradeJack acts as an autonomous, enterprise-grade procurement team. It constantly monitors unstructured global news, predicts supply shortages, secures voice-based authorization from the warehouse manager, and deploys a swarm of AI agents to physically navigate legacy (non-API) wholesaler websites to buy inventory at the lowest True Landed Cost.
-
🧠 Predict (The Macro-Event Trigger):
- Ingests real-time global news.
- Amazon Nova 2 Lite performs causal reasoning (e.g., "Heavy rains in Brazil" → "Coffee supply drops" → "Prices will spike").
-
📞 Authorize (The Gatekeeper):
- Triggers an emergency voice dispatch (Nova 2 Sonic architecture).
- Calls the deskless warehouse manager for a verbal "EXECUTE" command, bridging the physical and digital worlds.
-
🕵️ Verify (Anti-Monopoly Engine - TAME):
- Before purchasing, the AI calculates the markup against the raw commodity index.
- Uses economic proofs to detect and warn users of hidden monopolies artificially fixing prices.
- Formula:
$$Markup = \frac{\text{Landed Cost} - \text{Commodity Index Cost}}{\text{Commodity Index Cost}} \times 100$$
-
🤖 Execute (The UI Swarm):
- Deploys parallel browser agents (Nova Act architecture).
- Physically navigates clunky, legacy web portals, finds the "Add to Cart" buttons, and secures the cheapest inventory.
- Local Vendors & Retailers: Who need to secure inventory before wholesale prices spike.
- Deskless Warehouse Managers: Who need to authorize high-stakes purchases via voice while on the floor.
- Mid-Market Procurement Officers: Who want to avoid being price-gouged by hidden supplier monopolies.
This project utilizes a serverless, highly optimized architecture to ensure zero cost failures and rapid execution.
- Frontend: React + Vite (Custom dark-mode UI with portal scanline animations)
- AI Orchestration & Backend: Puter.js (Serverless framework)
- Reasoning Engine: Amazon Nova 2 Lite
- Security Layer: Delimiter Prompt Shielding & Token Truncation
- Caching & Persistence: Puter Key-Value Store (
puter.kv)
To ensure economic safety and prevent malicious use, TradeJack implements:
- Prompt Injection Security: Strict delimiter shielding prevents users from overriding the AI's core instructions.
- Cost Optimization: KV-Caching stores market reasoning for 24 hours, preventing redundant API calls and credit depletion.
- Economic Sanity Checks: Hardcoded UI validation prevents the execution of anomalous or hallucinated market prices.
- Node.js (v18+)
- A Puter.js account (Free - handles all AI authentication seamlessly)
- Clone the repository:
git clone [https://github.com/yourusername/tradejack.git](https://github.com/yourusername/tradejack.git) cd tradejack
2. **Install dependencies:**
```bash
npm install
- Run the development server:
npm run dev- Experience the Demo:
- Open
localhost:5173in your browser. - Paste a sample news headline (e.g., "Unexpected frost wipes out 40% of Florida's orange crop").
- Click Initiate Predictive Trigger and watch the autonomous swarm in action!
- Phase 2: Multimodal AI integration. Vendors will snap photos of physical delivery slips, and TradeJack will cross-reference them against digital receipts to flag unauthorized "unloading fees."
- Phase 3: Full AWS Fargate Docker integration for the live Nova Act browser swarm.