Clozyt Dev

• 

• 

  recommending with the filters

• 

  catalouge of the products

• 

  like the product it goes to database

Inspiration

From CSVs to AI Stylist in a Day

TL;DR: You turned two retail CSVs into a working AI clothes recommender with a Streamlit front end, optional vector search, and Gemini-powered styling tips—plus a plan to surface it in Glide.

What you set out to do

Make a hackathon-ready app that lets anyone type a natural-language request (e.g., “black oversized hoodie,” “summer brunch dress under $60”) and instantly get relevant products and practical styling advice.

What you actually built

Data → Catalog: Ingested two product datasets (ALTAR’D STATE and Alo Yoga), cleaned/standardized them, and produced a single combined_products_clean.csv (≈2K items total across both sources).

App UI: Wrote app_gemini.py (Streamlit). Users enter a style/theme; the app shows product cards (image, brand, price, link) and a “How to style these picks” section.

AI styling (Gemini): Wired google-generativeai so the app sends the top product images + theme to Gemini and returns 3–5 concise, actionable styling tips.

Search options:

Lite mode (no heavy deps): Keyword search over name/brand/category—simple and reliable.

Full mode (optional): Chroma + OpenCLIP for text→image semantic search; falls back to keyword if install isn’t available.

Dev-UX & Docs: Created one-click run scripts (Windows/Mac), a README, and a 90-second pitch. You also recorded demo videos.

Glide path (next step): Generated a FastAPI template exposing /search and /style, so Glide can consume JSON if/when you want a no-code mobile front end.

Your role & decisions

Builder & integrator: You chose speed over from-scratch ML, focusing on a clean UX, real catalog data, and an explainable AI layer.

Pragmatic architecture: Lite path ensures the demo always runs; vector search is an upgrade, not a blocker.

Judge-friendly narrative: Clear problem → working demo → AI value-add (styling tips) → deployment/Glide plan.

Outcome

A functioning AI clothes recommender that:

Answers natural-language shopping intents.

Surfaces real products with images and links.

Adds Gemini stylist tips that feel useful, not fluffy.

Reusable code and docs so you can ship locally, deploy, or hand it to a teammate.

What’s next (you already teed this up)

Deploy (Streamlit Cloud or small VM), set GEMINI_API_KEY as a secret.

Add Likes → CSV export and budget auto-detection (“under $60”).

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Clozyt Dev

Built With

• clip
• multimodular
• python