CELEBRITY MATCH

Inspiration

We wanted to create something fun, stylish, and interactive using just one AI prompt to show the power of what Gemini 2.0 Flash can do when guided with structure.
The idea of matching people with celebrities based on shared interests felt personal, exciting, and perfect for this constraint.

What it does

Celebrity Match is a full-stack web app where users enter three things they like (e.g., pizza, space, painting).
The app then calls Gemini 2.0 Flash with a structured, guardrailed prompt.
The AI responds with a Markdown-rendered table of celebrities who share those interests — complete with citations.

How we built it

🔥 Single Prompt Build

This project was built entirely from a single prompt using bolt.new for the One-Shot Challenge.

No edits were made to the code after the prompt was submitted.

Frontend

Built with React + Tailwind CSS
Styled in a Krico Retro Pop theme:
- Glassmorphism effects
- Gradient buttons
- 90s cartoon-style typography (Luckiest Guy)

Backend

Express.js server
Calls Gemini 2.0 Flash API using structured JSON format

Markdown Rendering

Uses marked.js to convert Gemini’s Markdown response into rich HTML

Security

Rate limiting for safe usage
Input validation
API key management using environment variables

Challenges we ran into

Prompt engineering was key crafting a prompt that consistently produced clean Markdown tables with citations was difficult.
Gemini sometimes hallucinated facts, so we forced structure using superscripts and source sections.
Dealing with CORS and API secrets while deploying a full-stack app.
Took 13 prompt attempts to make everything work correctly with zero manual edits.

Accomplishments we’re proud of

Built an entire Gemini-powered full-stack app from a single AI prompt — no manual coding.
Achieved a cohesive and retro-themed design using only prompt-based generation.
Got Gemini to generate accurate, source-cited celebrity match tables.

What we learned

With a well-structured prompt, it’s absolutely possible to go from idea to deployed app in a single step.
AI can now build not just mockups but complete, styled, functional apps when given clarity and constraints.