Skip to content

README.md

Latest commit

 

History

History
224 lines (174 loc) · 10.3 KB

README.md

File metadata and controls

224 lines (174 loc) · 10.3 KB
Raw

Stratis — Supply Chain Risk Intelligence Platform

A comprehensive supply chain risk management and visualization platform built for Unihack 2026. Stratis helps businesses understand, analyze, and mitigate geopolitical, logistical, and price risks across their global supply chains through interactive maps, AI-powered insights, and relocation simulations.

Next.js React TypeScript Tailwind CSS

Features

Interactive Supply Chain Map

  • Mapbox-powered global map with country and chokepoint visualization
  • Custom route builder with waypoints (origin, transit, destination)
  • Route finder for discovering supply paths between countries
  • Real-time risk overlay for countries and maritime chokepoints (Strait of Malacca, Suez Canal, Panama Canal, etc.)

AI-Powered Product Decomposition

  • Decomposes any product into a full supply chain dependency tree (4-5 tiers: product → subsystems → components → materials → geographies)
  • Generates 25-40 nodes with geographic concentration percentages, confidence scores, and risk factors
  • Evidence-backed verification via Perplexity Sonar real-time web search
  • Streams results via SSE for progressive UI updates

AI Risk Intelligence

  • Country risk evaluation — OpenAI GPT-4.1 Nano evaluates tariff, conflict, and policy risk factors for both import and export sides using structured JSON schema output
  • News-grounded scoring — Perplexity Sonar fetches real-time news headlines per country; the risk model uses these as evidence, falling back to base knowledge when headlines are irrelevant
  • Weighted risk calculation (tariff 35%, conflict 40%, policy 25%) with confidence scores and rationale per factor
  • Risk snapshots persisted to public/data/risk-snapshots.json
  • Price risk timeline and volatility visualization
  • Alert banners for critical supply chain disruptions

AI Supplier Alternatives & Optimization

  • Supplier alternatives — Kimi K2.5 (via OpenRouter) suggests 3-5 lower-risk sourcing countries for any high-risk component, filtered against a risk score database to guarantee safer results
  • Supply chain optimization — Kimi K2.5 analyzes a product's full supply chain tree and returns risk warnings, diversification strategies, and concrete optimization suggestions
  • Post-processing validates and normalizes all LLM outputs; deterministic fallback ensures results even without an API key

AI Relocation Analysis

  • Relocation analysis — Combines a deterministic scoring engine (market access, labor costs, infrastructure, trade agreements) with Kimi K2.5 qualitative enhancement for advantages, challenges, and geopolitical insights
  • Relocation simulation — Simulates moving production between any two countries with cost impact, timeline, chokepoint avoidance, and risk reduction projections
  • Country comparison — Side-by-side comparison of relocation targets

Supply Chain Insights

  • Health scoring and risk breakdown (geopolitical, logistics, price volatility)
  • Critical chokepoint exposure analysis
  • Actionable recommendations with potential savings
  • Route comparison and path details panel

AI Architecture

Stratis uses a multi-model architecture, choosing the right model for each task:

AI Provider Model Used For Why This Model
OpenAI GPT-4.1 Nano Country risk evaluation Strict JSON schema output, fast batch processing, low cost for structured scoring
OpenRouter Anthropic Claude Opus 4.6 (configurable) Product decomposition Deep reasoning for complex multi-tier supply chain trees
OpenRouter Moonshot Kimi K2.5 Alternatives, optimization, relocation analysis Strong generalist reasoning with good cost-quality balance
Perplexity Sonar Real-time news search, supply chain research Grounded web search for current events and evidence gathering
GDELT — News event data Broad geopolitical event monitoring

How AI Flows Work

  1. Risk Evaluation Pipeline: Perplexity Sonar fetches recent news per country → headlines are passed as evidence to GPT-4.1 Nano → structured risk scores (tariff/conflict/policy) are returned via JSON schema → weighted scores are computed server-side

  2. Product Decomposition Pipeline: User enters a product name → Claude Opus 4.6 generates a full dependency tree as JSON → Perplexity Sonar searches for evidence on each node → tree is refined with real-world production data → results stream to the client via SSE

  3. Supplier Alternatives Flow: User selects a high-risk component → Kimi K2.5 suggests lower-risk countries → server-side post-processing filters out countries above the risk threshold and deduplicates → deterministic fallback ensures results if the LLM returns poor suggestions

  4. Relocation Analysis Flow: User specifies industry, risk concerns, priorities, and target markets → deterministic engine scores candidates using trade agreements, labor costs, and infrastructure data → Kimi K2.5 enhances top recommendations with qualitative insights → graceful fallback to engine-only results if the AI call fails

Tech Stack

Category Technologies
Framework Next.js 16, React 19
Language TypeScript 5.7
Styling Tailwind CSS 4, shadcn/ui
Maps Mapbox GL, react-map-gl
Charts Recharts
AI Models OpenAI GPT-4.1 Nano, Anthropic Claude Opus 4.6, Moonshot Kimi K2.5, Perplexity Sonar
AI Routing OpenRouter, OpenAI API, Perplexity API
Data Sources GDELT (news events), Mapbox (geospatial)
Forms React Hook Form, Zod

Project Structure

├── app/
│   ├── page.tsx              # Main dashboard
│   ├── layout.tsx
│   ├── lib/                   # App-specific utilities (risk client, country utils)
│   └── api/                   # API routes
│       ├── risk-evaluate/     # OpenAI country risk evaluation
│       ├── ai/
│       │   ├── alternatives/  # Supplier alternatives (OpenRouter)
│       │   └── optimize/      # Optimization suggestions
│       ├── news/              # Perplexity news search
│       ├── gdelt/             # GDELT news integration
│       ├── decompose/        # Product decomposition pipeline
│       ├── relocation/
│       │   ├── simulate/      # Relocation simulation
│       │   ├── analyze/      # Relocation analysis
│       │   └── compare/      # Country comparison
│       └── save-risk-snapshots/
├── components/
│   ├── supply-chain-map.tsx   # Mapbox map + routes
│   ├── product-supply-chain.tsx
│   ├── risk-sidebar.tsx
│   ├── inventory-sidebar.tsx
│   ├── relocation-panel.tsx
│   ├── route-builder.tsx
│   ├── route-finder-panel.tsx
│   ├── path-details-panel.tsx
│   ├── supplier-recommendations.tsx
│   ├── price-risk-timeline.tsx
│   ├── alert-banner.tsx
│   └── ui/                   # shadcn components
├── lib/
│   ├── supply-chain-analyzer.ts
│   ├── route-finder.ts
│   ├── route-graph.ts
│   ├── risk-calculator.ts
│   ├── trade-agreements.ts
│   ├── relocation-engine.ts
│   ├── relocation-simulator.ts
│   └── decompose/            # Product decomposition (OpenRouter, Perplexity)
├── public/
│   └── data/
│       └── risk-snapshots.json
└── styles/
    └── globals.css

Getting Started

Prerequisites

  • Node.js 18+
  • npm or pnpm

Installation

# Clone the repository
git clone https://github.com/your-org/Unihack_2026_Stratis.git
cd Unihack_2026_Stratis

# Install dependencies
npm install --legacy-peer-deps

Environment Variables

Create a .env.local file in the project root:

# Required for map visualization
NEXT_PUBLIC_MAPBOX_ACCESS_TOKEN=your_mapbox_token

# Required for AI-powered features
OPENAI_API_KEY=your_openai_key
OPENROUTER_API_KEY=your_openrouter_key

# Optional: for news-based risk analysis
PERPLEXITY_API_KEY=your_perplexity_key

# Optional: for decomposition model override
OPENROUTER_MODEL=anthropic/claude-opus-4-6
Variable Required Description
NEXT_PUBLIC_MAPBOX_ACCESS_TOKEN Yes Mapbox access token for map display
OPENAI_API_KEY Yes OpenAI API key for country risk evaluation
OPENROUTER_API_KEY Yes OpenRouter key for alternatives, optimization, relocation
PERPLEXITY_API_KEY No Perplexity key for news search
OPENROUTER_MODEL No Override default model for decomposition

Run Development Server

npm run dev

Open http://localhost:3000 in your browser.

Build for Production

npm run build
npm start

API Routes

Route Method AI Model Description
/api/risk-evaluate POST GPT-4.1 Nano Batch country risk evaluation with structured JSON schema output
/api/ai/alternatives POST Kimi K2.5 Suggest lower-risk supplier alternatives with post-processing validation
/api/ai/optimize POST Kimi K2.5 Analyze supply chain tree and return optimization suggestions
/api/news GET Perplexity Sonar Fetch real-time news headlines per country for risk evidence
/api/gdelt GET — GDELT news event data integration
/api/decompose POST Claude Opus 4.6 + Perplexity Sonar Product decomposition with evidence search (SSE streaming)
/api/relocation/simulate POST Kimi K2.5 Simulate relocation with cost, timeline, and risk projections
/api/relocation/analyze POST Kimi K2.5 Relocation analysis with AI-enhanced qualitative insights
/api/relocation/compare POST — Deterministic country comparison for relocation
/api/save-risk-snapshots POST — Persist risk evaluation snapshots to disk

License

This project was built for Unihack 2026. See the repository for license details.