An automated tool built with Python that scrapes automotive listings from Mercado Libre (Argentina) and leverages Anthropic's Claude to clean, structure, and standardize the raw data into a clean JSON format.
- Python
- Playwright (Web Scraping & Browser Automation)
- Anthropic API (Claude 3.5 Sonnet) (Data Extraction & Natural Language Processing)
- JSON (Data Modeling)
- Web Scraping (
scraper.py):- Uses Playwright with custom browser headers, locales, and human-like delays (
slow_mo) to securely navigate and extract raw listing cards from Mercado Libre Argentina.
- Uses Playwright with custom browser headers, locales, and human-like delays (
- AI-Powered Data Processing (
procesador.py):- Sends the raw text data to Claude.
- Claude acts as an automotive market analyst, parsing unstructured strings into a strictly typed, uniform JSON array.
- Automatically handles currency normalization (estimating conversions to USD when necessary), flags misleading pricing/down-payments (
alerta_anticipo), and extracts clean model names, years, mileage, and links.
tracker-autos/
β
βββ scraper.py # Handles Playwright browser automation and raw data extraction
βββ procesador.py # Integrates with Anthropic API to parse and structure data into JSON
βββ README.md # Project documentation
βββ claude-key.txt # Paste your Claude API key here