Skip to content

kalaba992/Carinski-ekosistem-demo

Repository files navigation

Carinski Ekosistem — Public Demo

What This Is

Carinski Ekosistem is an AI-powered customs classification assistant that helps customs professionals, importers, and exporters determine Harmonized System (HS) codes for products.

The main project repository is private — it contains real customs data, production infrastructure configuration, audit systems, and operational documentation.

This repository is a sanitized public demo — a minimal technical showcase for external reviewers.

What This Demo Is

  • ✅ Guest-only browsing — no login required
  • ✅ Synthetic product examples (stainless steel screw, plastic cable tie, etc.)
  • ✅ Mock classification responses with illustrative HS codes
  • ✅ Confidence display, decision rationale, and validation steps
  • ✅ Mantine v8 UI components

What This Demo Is NOT

  • ❌ Not a production application
  • ❌ Not connected to any real backend (no Convex, no OpenAI, no Auth0)
  • ❌ Not providing legally binding customs classification
  • ❌ Not a replacement for official customs advice
  • ❌ Not containing real customs data, audit logs, or operational documents

Live Demo

https://carinski-ekosistem-demo.pages.dev

Tech Stack

Layer Technology
Framework React 18
Build Vite 5
Language TypeScript
UI Mantine v8
Icons Tabler Icons
Deployment Cloudflare Pages

Architecture Overview

User → ProductSelector → ChatArea → demoClassifier (mock)
                                        ↓
                              ClassificationResult
                              + ConfidenceBar
                              + AuditExplanation
  • All classification logic is client-side only
  • The mock classifier returns pre-defined illustrative responses
  • No network requests, no backend, no database
  • Every result is marked "Demo only — illustrative classification"

Security & Anti-Hallucination Concept

The full production system implements:

  1. Confidence scoring — multi-factor confidence calculation based on source reliability, GIR path completeness, and attribute match quality
  2. Decision rationale — structured audit output showing which GIR rules were applied, which chapter notes were consulted, and what evidence supports the classification
  3. Source validation — cross-referencing against official tariff databases
  4. Human-in-the-loop — all automated classifications require professional review before legal reliance

This demo visualizes the output format without the backend logic.

How to Help

Please do not only write "nice project." The most useful feedback is specific, critical, and actionable.

Choose one review lane and open a matching structured issue template:

Read the Review Guide before submitting.

License

Source-available sanitized demo. All rights reserved unless a LICENSE file is added.

About

Sanitized public demo of an AI-powered customs classification assistant

Resources

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors