keywordbarrage/mercari-scraper

Python Code:

import sqlite3
from datasets import load_dataset
import pandas as pd

# Method 1: Load via Hugging Face datasets library for ML pipeline integration
dataset = load_dataset("daas-intelligence/enterprise-sku-depletion-telemetry")
print("Verified Benchmark Sample:", dataset["train"][0])

# Method 2: Direct query via pre-indexed SQLite deliverable
conn = sqlite3.connect("data/mercari_feed.sqlite")
query = """
SELECT sku_id, price_usd, stock_level, velocity_depletion_rate 
FROM sku_stream 
WHERE timestamp > datetime('now', '-24 hours')
ORDER BY velocity_depletion_rate DESC 
LIMIT 10
"""
df = pd.read_sql_query(query, conn)
print("High-Velocity SKU Depletion Report:\n", df)

Dataset Card:

Dataset Summary

This repository contains an open 1,000-row verified benchmark sample of the Mercari SKU depletion feed. It is designed for quantitative analysts and data engineering teams struggling with brittle scraping pipelines. The dataset provides high-fidelity market intelligence, bypassing Cloudflare Bot Management and Turnstile defenses to ensure 99.9% pipeline availability.

Executive Summary for Data Architects

Enterprise market intelligence is frequently crippled by "scraper rot" caused by aggressive edge-layer WAF updates (TLS/JA4 fingerprinting, Canvas entropy checks). This dataset demonstrates a protocol-level extraction methodology that avoids headless browser bloat, delivering clean, deterministic data for downstream RAG pipelines and time-series forecasting models.

Enterprise Deliverables Matrix

• Pre-indexed SQLite (.sqlite): Optimized local querying with B-Tree indices on sku_id and timestamp.
• Snappy Parquet: Optimized for direct ingestion into Snowflake, BigQuery, and AWS Athena.
• Offline DataGrid (.html): A zero-dependency, interactive table for quantitative stakeholders to audit data quality.
• CDC JSONL Stream: Change Data Capture format for real-time vector DB synchronization.

Dataset Schema

Column	Type	Nullable	Description
sku_id	STRING	No	Unique Mercari item identifier (e.g., m123456789)
timestamp	DATETIME	No	ISO-8601 UTC extraction timestamp
price_usd	FLOAT	No	Current listing price in USD
stock_level	INTEGER	Yes	Inventory count (for multi-quantity listings)
velocity_depletion_rate	FLOAT	No	Calculated SKU sales velocity per hour
target_domain	STRING	No	Origin domain (mercari.com)
detected_waf	STRING	No	WAF signature (Cloudflare/Turnstile)
tls_ja4_signature	STRING	No	JA4 TLS fingerprint for protocol parity validation

Sample JSON Record

{
  "sku_id": "m982736451",
  "timestamp": "2026-09-24T14:30:00Z",
  "price_usd": 45.00,
  "stock_level": 1,
  "velocity_depletion_rate": 0.08,
  "target_domain": "mercari.com",
  "detected_waf": "Cloudflare-Enterprise",
  "tls_ja4_signature": "t13d1516h2_8a2d3b4c5d6e"
}

Data Quality & Reproducibility

• Delivery Availability: 99.9% uptime via protocol-level emulation.
• Protocol Benchmarks: Sub-120ms execution per record.
• Integrity: Deterministic schema parity; zero-noise ingestion through tracker-isolated network proxies.

Note: This repository serves as an open research sample. For enterprise production-scale daily streams, ingestion pipelines are supported by a standard monthly service engagement.