keywordbarrage/rozetka-scraper

Python Code:

import sqlite3
from datasets import load_dataset
import pandas as pd

# Method 1: Load via Hugging Face datasets library
# This repository contains a verified benchmark sample.
# For full enterprise stream access, refer to the "Commercial Terms" section.
dataset = load_dataset("daas-intelligence/enterprise-rozetka-sku-depletion-and-waf-telemetry")
print("Hub Sample Record:", dataset["train"][0])

# Method 2: Load zero-latency local pre-indexed SQLite deliverable
# Assuming 'enterprise_rozetka_feed.sqlite' is available locally from a full deliverable.
# This sample demonstrates querying a pre-indexed SQLite dataset.
try:
    conn = sqlite3.connect("enterprise_rozetka_feed.sqlite")
    df = pd.read_sql_query("SELECT sku_id, price_uah, stock_level, depletion_velocity FROM sku_stream ORDER BY timestamp DESC LIMIT 10", conn)
    print("\nSQLite Sample Data:")
    print(df)
    conn.close()
except sqlite3.Error as e:
    print(f"\nCould not connect to SQLite database or execute query: {e}")
    print("Ensure 'enterprise_rozetka_feed.sqlite' is available locally. This is a benchmark sample.")

Dataset Card:

Enterprise High-Frequency SKU Depletion & WAF Telemetry Feed: Rozetka

This repository contains an open 1,000-row verified benchmark sample of the Rozetka (rozetka.com.ua) SKU depletion and WAF telemetry feed, structured with pre-indexed SQLite and Snappy Parquet partitions for quantitative market intelligence research and data engineering evaluation.

Dataset Summary & Research Motivation

Quantitative analysts, data scientists, and ML engineering teams often struggle with the fragility of traditional web scraping pipelines. Brittle scrapers, rotating proxy pools, and escalating anti-bot defenses (such as Cloudflare Bot Management or Enterprise Turnstile) lead to unstable data feeds, data loss, and significant operational overhead. This dataset addresses these challenges by providing a robust, protocol-level data extraction benchmark from Rozetka, Ukraine's leading e-commerce marketplace.

Our motivation is to provide a reliable, high-fidelity data source for understanding e-commerce dynamics, competitive pricing, and inventory movements on platforms like Rozetka, even in the face of advanced Web Application Firewalls (WAFs). This benchmark sample demonstrates our ability to bypass sophisticated edge challenge loops, TLS/JA4 fingerprinting, and behavioral entropy analysis with 99.9% verified pipeline availability, delivering clean, structured data ready for immediate analysis.

Zero-Code Enterprise Deliverables Matrix

This dataset showcases the four enterprise-grade formats provided by our fully-managed data extraction service, designed for seamless integration into diverse data architectures:

1. Pre-indexed SQLite (.sqlite): Instant local querying with pre-built indices on critical fields such as sku_id, price_uah, stock_level, depletion_velocity, and timestamp. Ideal for rapid prototyping, local development, and small-to-medium scale analytical tasks without requiring complex database setup.
2. Snappy Parquet: Columnar storage format optimized for efficient ingestion into cloud data warehouses like Snowflake, Google BigQuery, and AWS Athena. Parquet files are compressed with Snappy for optimal read performance and reduced storage costs, making them suitable for large-scale data lake environments.
3. Zero-Dependency Offline DataGrid (.html): A self-contained, interactive HTML table viewer. This deliverable allows quantitative stakeholders, business intelligence teams, and non-technical users to explore, filter, and sort the dataset directly in a web browser, requiring no additional software, database connections, or API keys.
4. CDC JSONL Stream: Change Data Capture (CDC) formatted JSON Lines (JSONL) for efficient updates. This stream is optimized for enterprise RAG (Retrieval Augmented Generation) pipelines, vector database updates, and real-time data synchronization, ensuring that downstream systems always have the freshest data with minimal latency.

Dataset Schema & Data Dictionary

The default configuration of this dataset includes the following columns:

Column Name	Type	Nullable	Description
sku_id	STRING	NO	Unique identifier for the product/SKU on Rozetka (Artikul / Kod tovaru).
timestamp	TIMESTAMP	NO	UTC timestamp of the data extraction.
product_title	STRING	NO	Full title of the product.
brand	STRING	YES	Brand or manufacturer of the product (e.g., Apple, Samsung, Bosch).
category_hierarchy	STRING	NO	Full marketplace category path (e.g., Elektronika > Smartfony > Mobilni Telefony).
price_uah	FLOAT	NO	Current retail price in Ukrainian Hryvnia (UAH).
original_price_uah	FLOAT	YES	Original crossed-out price in UAH, if a promotional discount is active.
discount_delta_uah	FLOAT	YES	Monetary value of the discount (Original Price - Current Price).
seller_type	STRING	NO	Classification of the seller: "Rozetka" (Prodavets Rozetka) or "Marketplace Merchant" (Inshyj prodavets).
stock_level	STRING	NO	Current stock availability (e.g., "In Stock", "Out of Stock", "Limited Stock").
depletion_velocity	FLOAT	YES	Calculated rate of SKU inventory depletion over time (for time-series data).
rozetka_premium_perks	BOOLEAN	NO	Indicates if Rozetka Premium loyalty perks (e.g., free delivery) apply.
fulfillment_modes	STRING	YES	Available fulfillment options (e.g., "Free Pick-up from Rozetka Stores", "Nova Poshta", "Courier").
rating_stars	FLOAT	YES	Average customer review rating out of 5 stars.
total_reviews	INTEGER	YES	Total number of verified buyer reviews.
technical_specs_json	STRING	YES	JSON string containing key-value pairs of detailed technical specifications (Kharakterystyky tovara).
target_domain	STRING	NO	The target domain from which data was extracted (e.g., rozetka.com.ua).
detected_waf	STRING	NO	Detected Web Application Firewall or Anti-Bot solution (e.g., Cloudflare Bot Management).
tls_ja4_signature	STRING	YES	The TLS/JA4 fingerprint of the client used for extraction, indicating protocol-level emulation.
delivery_status	STRING	NO	Status of the data point delivery (SUCCESS, WAF_BYPASS, ERROR).

Sample JSON Record

{
  "sku_id": "87569123",
  "timestamp": "2026-09-24T14:30:00Z",
  "product_title": "Смартфон Apple iPhone 15 Pro Max 256GB Natural Titanium",
  "brand": "Apple",
  "category_hierarchy": "Електроніка > Смартфони, ТВ та Електроніка > Смартфони",
  "price_uah": 52999.00,
  "original_price_uah": 54999.00,
  "discount_delta_uah": 2000.00,
  "seller_type": "Rozetka",
  "stock_level": "In Stock",
  "depletion_velocity": 0.05,
  "rozetka_premium_perks": true,
  "fulfillment_modes": "Free Pick-up from Rozetka Stores, Nova Poshta, Courier",
  "rating_stars": 4.8,
  "total_reviews": 1250,
  "technical_specs_json": "{\"Model\": \"iPhone 15 Pro Max\", \"Колір\": \"Natural Titanium\", \"Об'єм пам'яті\": \"256GB\", \"Гарантія\": \"12 місяців\"}",
  "target_domain": "rozetka.com.ua",
  "detected_waf": "Cloudflare Bot Management",
  "tls_ja4_signature": "t13,0-23-65281-10-11-35-16-5-13-18-51-45-43-21-47-49-39-50-52-19-20-53-54-25-27-28-29-30-31-32-33-34-36-37"