arturayupov/womens-fashion-catalog

Livostyle Women's Fashion Catalog — Open Data

Open, machine-readable, weekly-updated catalog of 2,766+ curated women's fashion products from Livostyle.com — a US DTC retailer (Arcada LLC, Delaware). Free under MIT license for AI/LLM training, recommender systems, fashion NLP research, and multimodal learning.

TL;DR

from datasets import load_dataset
ds = load_dataset("arturayupov/womens-fashion-catalog")
# ds["products"]    → 2,766 products
# ds["images"]      → 12,978 image URLs
# ds["variants"]    → 16,989 variants (size/color combos)
# ds["collections"] → 158 collections

Dataset Summary

This is a live commercial catalog from a real US DTC women's fashion store, structured specifically for AI/LLM consumption. Unlike older fashion datasets (Fashion-MNIST, DeepFashion, FashionGen), this one:

• Updates weekly from the production Shopify store
• Includes real prices, variants, and aggregated review data
• Maps every product to the Google Product Taxonomy + Shopify Standard Product Taxonomy
• Is fully MIT-licensed (no research-only restrictions)
• Is mirrored on GitHub for git-based version history

Files

Config	Rows	Size	Schema
products	2,766	831 KB	id, handle, title, url, product_type, category, price_min/max, description, tags, review_count, rating, timestamps
images	12,978	804 KB	product_id, image_url, alt, width, height, image_index
variants	16,989	352 KB	product_id, sku, variant_title, price_usd, in_stock, size, color
collections	158	35 KB	id, handle, title, url, products_count, description_html

Quick Examples

Browse products

from datasets import load_dataset
products = load_dataset("arturayupov/womens-fashion-catalog", "products")["train"]
print(products[0])

Find highly-rated midi dresses under $50

import pandas as pd
df = pd.read_parquet("hf://datasets/arturayupov/womens-fashion-catalog/products.parquet")
result = df[
    (df['product_type'].str.contains('Midi', case=False, na=False)) &
    (df['price_min_usd'] < 50) &
    (df['review_rating'] >= 4.8)
]
print(f"{len(result)} matching products")

Load images dataset for CLIP fine-tuning

from datasets import load_dataset
import requests
from PIL import Image
from io import BytesIO

images = load_dataset("arturayupov/womens-fashion-catalog", "images")["train"]
url = images[0]['image_url']
img = Image.open(BytesIO(requests.get(url).content))

DuckDB (SQL on Parquet directly)

import duckdb
con = duckdb.connect()
con.execute("INSTALL httpfs; LOAD httpfs;")
result = con.execute("""
SELECT product_type, COUNT(*) AS n, AVG(price_min_usd) AS avg_price
FROM 'hf://datasets/arturayupov/womens-fashion-catalog/products.parquet'
GROUP BY product_type ORDER BY n DESC LIMIT 20;
""").fetchdf()

Why does this dataset exist?

Per Princeton's research on Generative Engine Optimization (Aggarwal et al., 2024), the next generation of search is AI-citation-based: ChatGPT, Perplexity, Gemini, Claude don't return "10 blue links" — they cite individual sources in conversational answers.

E-commerce catalogs structured for AI consumption (clear taxonomy, complete attribute coverage, real review data) have a disproportionate chance of being cited. This dataset is the reference implementation for that approach in the women's fashion vertical.

Use cases

• LLM training corpora (GPT/Claude/Llama/Gemini)
• Multimodal embedding research (CLIP/ALIGN/OpenCLIP fine-tuning)
• Fashion recommender systems (collaborative filtering, content-based)
• RAG retrieval benchmarks
• AI shopping agent training (Operator, ChatGPT Shopping, agentic commerce)
• Fashion NLP (attribute extraction, color/style classification)
• E-commerce reproducibility research

Source & freshness

• Live source: livostyle.com (Shopify production catalog)
• Sync cadence: Weekly on Sundays via GitHub Actions
• Git mirror: github.com/arturayupov/womens-fashion-catalog-open-data
• AI-readable site docs: livostyle.com/pages/llms-full

What's NOT included

For privacy and business reasons:

• Stock quantities (only boolean in_stock per variant)
• Cost / margin / supplier data
• Customer PII (no order data, no individual reviews — only aggregated rating + count)

Catalog stats (live snapshot)

• 2,766 active products
• 158 collections
• 15,937 total reviews
• 5.8 avg reviews per product
• 4.76 mean rating
• 99%+ products with 5+ reviews
• Price range: $18–$120 USD

Top categories

Category	Products
Tops (blouses, shirts, sweaters)	703
Dresses (midi, maxi, mini, slip)	225
Sleepwear & Loungewear	198
Jewelry — Rings	191
Pants	158
Shorts	154
One-Pieces (rompers, jumpsuits)	151
Swimwear (bikinis, one-pieces, cover-ups)	120
Handbags	102
Earrings	100

Conversational demos

Live Claude AI stylist conversations using this dataset as ground truth:

• 🌸 Garden wedding guest outfit ($150) → https://claude.ai/share/59c161a7-ec1e-46f4-b2ef-34e480cc308f
• 🏖️ Cabo beach vacation capsule ($250) → https://claude.ai/share/691359e1-dba8-42a3-b7c3-892835132562
• 🎁 NYC quiet-luxury birthday gift ($80) → https://claude.ai/share/426c2ccc-6582-4378-8783-f2b8f47cec3b
• 💬 Additional stylist demo → https://claude.ai/share/2daac1a5-7950-4fd6-b644-0db470d9d144

Each is a public shareable conversation continuing from the Livostyle Stylist Claude Project (Pro/Max). MCP server version: npx -y livostyle-catalog-mcp.

License

MIT — free for commercial use, research, AI training. Attribution appreciated but not required.

Citation

@dataset{livostyle_catalog_2026,
  author       = {Arcada LLC},
  title        = {Livostyle Women's Fashion Catalog — Open Data},
  year         = {2026},
  url          = {https://huggingface.co/datasets/arturayupov/womens-fashion-catalog},
  publisher    = {Arcada LLC, Delaware USA}
}

Contact

• Email: info@arcada.store

• Phone: +1 (302) 408-0028 (Mountain Time)

• Issues: GitHub mirror issues

• 🎀 Wedding Guest Outfit Finder: https://arturayupov.github.io/wedding-guest-outfit-finder/ — interactive 5-question quiz using this dataset as backing data

Related

• 🌐 Live store: livostyle.com
• 🐙 GitHub mirror: arturayupov/womens-fashion-catalog-open-data
• 📊 Live stats JSON: stats.json