Unboxd

Creating a listing - powered by AI
Trade Suggestions
Spotlight!
Mainpage
Stuck in books
Engrossed in knowledge

## Inspiration

We are both avid figurine collectors and often ran into the problem of opening duplicates, which sometimes left us with figurines we didn’t want. We wanted a way to give our figurines a new purpose and make trading easier and more fair.

What it does

Unboxd is a trading platform for figurine collectors. Users type the name of a figurine they want to list, and the platform automatically fills in details like price and type using the Gemini API. Then it matches users with compatible figurines based on value and personal preferences, making trades faster, fairer, and more fun.

How we built it

We built a full-stack platform using:

Languages: JavaScript (Node.js backend, React frontend)
Frontend: React (JSX), Vite (dev server/bundler)
Backend: Node.js, Express, dotenv, jsonwebtoken (JWT auth)
AI/ML: Google Gemini API for value and type inference
API style: REST endpoints (/auth, /listings, /estimate)
Database: PostgreSQL (DATABASE_URL in .env)
Tooling: npm, VS Code

Challenges we ran into

Reliable AI estimation: Forcing Gemini to return structured JSON and handling variability, plus creating local fallbacks when needed.
UX flow changes: Supporting name-only estimation and reflecting value ranges accurately.
Authentication robustness: Handling expired tokens and server restarts with clear error messages and auto-redirects.
Development networking: Fixing port conflicts and ensuring .env loaded correctly.
Data consistency: Relaxing validation while reliably inferring type and brand.
Demo reliability: Seeding data and making the end-to-end flow resilient with helpful logs.

Accomplishments that we're proud of

We successfully integrated AI to automate value and type inference, built a fully functional trading platform with end-to-end flow, and made it user-friendly for collectors even without web development experience.

What we learned

We learned how to navigate GitHub, use VS Code efficiently, integrate APIs, handle real-world data inconsistencies, and build both frontend and backend systems. We also gained valuable experience in problem-solving, UX design, and full-stack development.

What's next for Unboxd

We plan to improve AI estimation accuracy, enhance matching algorithms, expand the user base, and explore additional features like community trading events and collectible analytics.