ToyotaTinder

AI ranked Toyota matchmaker with swipe style discovery for Hackathon 2025.

ToyotaTinder lets shoppers enter budget and lifestyle preferences, then uses a Toyota inventory CSV plus Google Gemini 2.5 Flash to surface best fit cars for swiping, saving, and revisiting.

---

Features

• Preference intake (/find) Simple form for budget, mileage, fuel type, use case, and notes.

• AI ranking (/api/analyze) Sends up to 75 filtered rows from CarData.csv to Gemini 2.5 Flash and expects JSON with selected IDs and reasoning. If Gemini fails or no API key is set, falls back to a local scoring function or random but clearly labeled matches.

• Swipe deck (/swipe) Stacked, draggable card deck built with Framer Motion plus click buttons for like or skip.

• Persistent likes and results Matches, AI notes, and liked cars are stored in localStorage so refreshes and navigation do not lose progress.

• Lightweight auth /signup and /login store accounts and a simple session token in localStorage. AuthGate protects swipe and liked routes. This is for demo only and not secure for production.

• Liked garage (/liked) Shows all liked cars and gives a clear path to restart matching or clear saved likes.

---

Architecture

Layer	Details
UI	Next.js App Router with React 19, Tailwind CSS utilities in src/app/globals.css, Framer Motion animations, simple page transition wrapper.
Data	src/data/CarData.csv loaded on the server and converted into typed Car objects.
AI	@google/genai client calls Gemini 2.5 Flash with strict JSON output, with local scoring fallback.
State	likes.ts wraps localStorage for likes and results, AuthGate checks a tt-auth token.

Matching Flow

1. User fills in preferences on /find.
2. Server filters the CSV by budget, fuel type, mileage, and optional location.
3. Up to 75 rows are sent to Gemini 2.5 Flash with responseMimeType: application/json.
4. Gemini returns car IDs plus short reasoning.
5. IDs map back to Car objects and are saved for the client.
6. /swipe shows the deck and records likes into a saved list used by /liked.
7. If Gemini response is missing or invalid, the app logs a warning and uses local scoring or random selection so the demo still works.

Setup

Requirements: Node.js 20 or higher, npm 10 or higher, optional Gemini API key.

npm install
npm run dev       # http://localhost:3000

Environment variables in .env.local:

GEMINI_API_KEY=your-google-genai-key

Without a key, matching still runs using the local heuristic or random fallback with a visible warning.

Useful scripts:

• npm run dev local development
• npm run build production build
• npm run start run built app
• npm run lint run ESLint

---