keywordbarrage/carousell-scraper

Python Code:

import sqlite3
from datasets import load_dataset
import pandas as pd

# Method 1: Load via Hugging Face datasets library
# This loads a benchmark sample, actual enterprise feeds are delivered via direct streams.
dataset = load_dataset("daas-intelligence/carousell-recommerce-sku-depletion-and-waf-telemetry")
print("Hub Sample Record:", dataset["train"][0])

# Method 2: Load zero-latency local pre-indexed SQLite deliverable
# This demonstrates loading the production-grade SQLite deliverable.
# Replace "carousell_feed.sqlite" with your actual file path.
try:
    conn = sqlite3.connect("carousell_feed.sqlite")
    df = pd.read_sql_query("""
        SELECT listing_id, listing_title, listing_price_sgd, item_condition, 
               seller_username, deal_status, timestamp 
        FROM carousell_listings 
        ORDER BY timestamp DESC 
        LIMIT 10
    """, conn)
    print("\nSQLite Sample Data (first 10 records):")
    print(df)
    conn.close()
except sqlite3.OperationalError:
    print("\n'carousell_feed.sqlite' not found. Please ensure the SQLite deliverable is available locally.")

Dataset Card:

---

pretty_name: "Enterprise High-Frequency SKU Depletion & WAF Telemetry Feed: Carousell" annotations_creators: - machine-generated language: - en - tr license: apache-2.0 task_categories: - tabular-classification - time-series-forecasting tags: - carousell-recommerce - sku-depletion-data - waf-bypass-telemetry - time-series-forecasting - e-commerce-intelligence - sqlite-parquet-delivery size_categories: - 10K<n<100K configs: - config_name: default data_files: - split: train path: "data/train-.parquet" - split: validation path: "data/val-.parquet"

Carousell Recommerce SKU Depletion & WAF Telemetry Dataset

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

Dataset Summary & Research Motivation

The Carousell Recommerce SKU Depletion & WAF Telemetry dataset provides production-grade, high-frequency market intelligence from carousell.sg, a leading Southeast Asian classifieds and recommerce platform. For quantitative analysts and data engineering teams, maintaining reliable data pipelines from dynamic, WAF-hardened targets like Carousell is a persistent challenge. Brittle scrapers, rotating CAPTCHAs, and broken residential proxy pools frequently lead to HTTP 403 bans, TLS/JA4 fingerprinting blocks, and ultimately, destroyed data pipelines.

This dataset addresses these critical enterprise pain points by demonstrating protocol-level data extraction, bypassing sophisticated edge challenge loops, including Cloudflare Bot Management, Enterprise Turnstile, and Envoy Gateway defenses. Our managed ingestion pipelines ensure 99.9% verified data availability, delivering clean, structured data streams for tracking real-time inventory movement, pricing elasticity, seller credibility, and SKU depletion velocity without the overhead of maintaining internal scraping infrastructure.

Zero-Code Enterprise Deliverables Matrix

This dataset is delivered in four enterprise-ready formats, designed for seamless integration into diverse quantitative research and data engineering workflows:

• Pre-indexed SQLite (.sqlite): Instant local querying with pre-built indices on listing_id, timestamp, listing_price_sgd, and deal_status. Ideal for rapid prototyping, ad-hoc analysis, and local BI dashboards.
• Snappy Parquet: Optimized for direct ingestion into cloud data warehouses such as Snowflake, Google BigQuery, and AWS Athena. Parquet offers columnar storage efficiency and Snappy compression for reduced storage costs and faster query performance.
• Zero-Dependency Offline DataGrid (.html): A self-contained, interactive HTML table viewer for quantitative stakeholders and business users. This deliverable requires no SQL knowledge, API keys, or SaaS subscriptions, providing a portable, shareable view of the data.
• CDC JSONL Stream: Change Data Capture (CDC) formatted JSON Lines, perfect for real-time updates to Enterprise RAG pipelines, vector databases, and event-driven architectures. Each line represents a record with metadata indicating changes.

Dataset Schema & Data Dictionary

The carousell_listings table contains the following fields:

Column Name	Type	Nullable	Description
listing_id	TEXT	NO	Unique identifier for the Carousell listing (e.g., ItemId / Slug).
canonical_url	TEXT	NO	The canonical URL of the listing page.
timestamp	TIMESTAMP	NO	UTC timestamp of when the data record was extracted.
listing_title	TEXT	NO	The full title of the product listing.
brand	TEXT	YES	Brand name of the item.
model	TEXT	YES	Model name or identifier of the item.
item_condition	TEXT	NO	Condition grade of the item (e.g., 'Brand New', 'Like New', 'Lightly Used', 'Well Used').
category_hierarchy	TEXT	NO	Hierarchical category path of the listing (e.g., 'Electronics > Mobile Phones').
listing_price_sgd	REAL	NO	Listing price in Singapore Dollars (SGD). Other currencies may appear in listing_price_raw.
listing_price_raw	TEXT	NO	Raw listing price string, including currency symbol (e.g., 'S$150', 'HKD 750').
price_negotiable	BOOLEAN	NO	Indicates if the price is negotiable (True/False).
seller_username	TEXT	NO	Username of the seller.
seller_rating	REAL	YES	Average star rating of the seller (out of 5).
total_reviews	INTEGER	YES	Total number of reviews for the seller.
seller_joined_date	TEXT	YES	Date when the seller joined Carousell.
singpass_verified	BOOLEAN	NO	Indicates if the seller is Singpass verified (True/False).
caroubiz_status	BOOLEAN	NO	Indicates if the seller has a CarouBiz merchant tier subscription (True/False).
carousell_protection_eligible	BOOLEAN	NO	Indicates if the listing is eligible for Carousell Protection escrow payment (True/False).
deal_method	TEXT	NO	Preferred deal method (e.g., 'Meetup', 'Mail', 'Courier').
meetup_location	TEXT	YES	Geotagged meetup location or MRT station (e.g., 'Jurong East MRT').
shipping_fee_sgd	REAL	YES	Shipping fee in SGD, if applicable.
date_posted	TEXT	NO	Original date the listing was posted (e.g., '2 days ago', '24 Sep').
likes_count	INTEGER	NO	Number of likes/favorites the listing has received.
deal_status	TEXT	NO	Current status of the deal (e.g., 'Active', 'Reserved', 'Sold').
seller_response_rate	TEXT	YES	Seller's chat response rate percentage (e.g., '95%').
average_reply_time	TEXT	YES	Seller's average reply time (e.g., 'Within minutes', 'Within an hour').
detected_waf	TEXT	NO	Detected Web Application Firewall (e.g., 'Cloudflare Bot Management').
tls_ja4_signature	TEXT	NO	The TLS/JA4 fingerprint signature observed during extraction.
delivery_status	TEXT	NO	Status of data delivery for this record ('SUCCESS', 'FAILURE_WAF_CHALLENGE', etc.).

Sample JSON Record

{
  "listing_id": "carousell-5f2b8e3a-e9c7-4d1a-b0f3-8c7e6a1b2d3f",
  "canonical_url": "https://www.carousell.sg/p/apple-iphone-14-pro-max-128gb-123456789",
  "timestamp": "2026-09-24T14:30:00Z",
  "listing_title": "Apple iPhone 14 Pro Max 128GB - Deep Purple (Like New)",
  "brand": "Apple",
  "model": "iPhone 14 Pro Max",
  "item_condition": "Like New",
  "category_hierarchy": "Electronics > Mobile Phones > iOS Phones",
  "listing_price_sgd": 1250.00,
  "listing_price_raw": "S$1,250",
  "price_negotiable": true,
  "seller_username": "TechDealsSG",
  "seller_rating": 4.8,
  "total_reviews": 312,
  "seller_joined_date": "2022-01-15",
  "singpass_verified": true,
  "caroubiz_status": false,
  "carousell_protection_eligible": true,
  "deal_method": "Meetup",
  "meetup_location": "Orchard MRT Station",
  "shipping_fee_sgd": null,
  "date_posted": "1 day ago",
  "likes_count": 87,
  "deal_status": "Active",
  "seller_response_rate": "98%",
  "average_reply_time": "Within minutes",
  "detected_waf": "Cloudflare Bot Management",
  "tls_ja4_signature": "t13d1600000000000000000000000000000000000000000000000000000000000000",
  "delivery_status": "SUCCESS"
}

Python Usage Code

import sqlite3
from datasets import load_dataset
import pandas as pd

# Method 1: Load via Hugging Face datasets library
# This loads a benchmark sample, actual enterprise feeds are delivered via direct streams.
dataset = load_dataset("daas-intelligence/carousell-recommerce-sku-depletion-and-waf-telemetry")
print("Hub Sample Record:", dataset["train"][0])

# Method 2: Load zero-latency local pre-indexed SQLite deliverable
# This demonstrates loading the production-grade SQLite deliverable.
# Replace "carousell_feed.sqlite" with your actual file path.
try:
    conn = sqlite3.connect("carousell_feed.sqlite")
    df = pd.read_sql_query("""
        SELECT listing_id, listing_title, listing_price_sgd, item_condition, 
               seller_username, deal_status, timestamp 
        FROM carousell_listings 
        ORDER BY timestamp DESC 
        LIMIT 10
    """, conn)
    print("\nSQLite Sample Data (first 10 records):")
    print(df)
    conn.close()
except sqlite3.OperationalError:
    print("\n'carousell_feed.sqlite' not found. Please ensure the SQLite deliverable is available locally.")

Reproducibility & Data Quality Metrics

Our enterprise data pipelines are engineered for maximum reliability and data integrity against hardened web application firewalls (WAFs) and advanced bot mitigation systems.

• 99.9% Delivery Availability: Guaranteed uptime for continuous data ingestion from Carousell, ensuring consistent access to critical market intelligence.
• Deterministic Schema Parity: Strict schema validation and normalization ensure that every delivered dataset conforms precisely to the defined structure, preventing schema drift and enabling seamless integration.
• Sub-120ms Protocol Execution: Our low-level protocol-based extraction engines achieve sub-120ms execution times per request, minimizing latency and maximizing throughput for high-frequency data streams.
• WAF Bypass & Telemetry Validation: Each data point is extracted with verified WAF bypass success, including comprehensive telemetry validation for TLS/JA4 signatures, HTTP/3 fingerprinting, and behavioral consistency, ensuring the authenticity and completeness of the data.

This repository offers a benchmark sample, demonstrating the quality and structure of our enterprise-grade data feeds. For ongoing access to the full Carousell SKU depletion feed, including real-time updates and custom schema requirements, a minimum enterprise engagement starts at 1,500 USDT per month. This retainer covers the fully managed pipeline, WAF bypass, infrastructure, and direct data delivery.