Looking for critical feedback on a sanitized public demo of my AI customs classification assistant

[kalaba992]

🏷️ Discussion Type

Question

Body

Hi everyone,

I’m Edin, and I work in customs/import-export.

I’m building an AI-powered customs classification assistant for HS code determination, auditability, and anti-hallucination validation.

The main project is private, so I created a sanitized public demo repo and a live guest-only demo app for review.

Public demo repo:
https://github.com/kalaba992/Carinski-ekosistem-demo

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

The demo contains no private data, no credentials, no Auth0, no Convex, no OpenAI, and no production infrastructure. It uses synthetic examples and illustrative classification outputs only.

I’m looking for direct, critical feedback in these lanes:

1. Architecture review
2. Security review
3. Testing strategy
4. UI/UX feedback
5. Customs-domain / legal wording review
6. Bug report

Review guide:
https://github.com/kalaba992/Carinski-ekosistem-demo/blob/main/docs/REVIEW_GUIDE.md

The questions I care about most:

• What would make this unsafe to trust?
• What would block this from becoming production-grade?
• What architectural decision should be changed early?
• What security risk would be expensive to fix later?
• What wording creates legal or professional overclaim risk?
• What test would give the most confidence?

I’m not asking anyone to build it for me. I’m looking for honest technical review and experienced eyes on weak spots.

Thank you.

Guidelines

• I have read and understood this category's guidelines before making this post.