ecommerce-price-watcher

# Ecommerce Price Watcher Monitor product URLs, keep price history, detect offers, and output alert-ready JSON. ## Quick start Use `scripts/price_watch.py`. ```bash # URL mode python3 skills/ecommerce-price-watcher/scripts/price_watch.py add \ --url "https://example.com/product" \ --target-price 399990 \ --currency CLP # Item mode (discover URLs from query) python3 skills/ecommerce-price-watcher/scripts/price_watch.py add-item \ --query "iPhone 13 128GB Chile" \ --target-price 349990 \ --currency CLP \ --trusted-only \ --max-results 5 python3 skills/ecommerce-price-watcher/scripts/price_watch.py check --all ``` ## Commands - `add`: add a single product URL - `add-item`: discover product URLs from an item query, then add watches - `list`: list watched products - `check --id <id>`: check one product now - `check --all`: check all products now - `remove --id <id>`: delete watcher - `history --id <id>`: print full price history ## Alert behavior A check produces alerts when at least one condition matches: 1. `price_drop`: current price < previous price 2. `target_hit`: current price <= target price Alert payload includes: - product id - title - old/new price - drop percent (when available) - URL - timestamp ## Item-query mode details `add-item` uses a lightweight search discovery flow to find candidate product links. - `--trusted-only` restricts discovered URLs to a curated trusted domain list. - `--max-results` controls how many links are added. - Duplicate URLs are skipped safely. This gives users natural language entry ("track iPhone 13 128GB") instead of forcing direct URLs. ## Parsing strategy Use a layered parser: 1. JSON-LD `offers.price` 2. Open Graph/meta price fields 3. Generic HTML regex fallback When multiple prices are found, choose the lowest positive value as the current offer candidate. ## Security standards - Accept only `http`/`https` URLs. - Enforce request timeout. - Enforce response body size cap. - Do not execute remote JavaScript. - Store no API keys/tokens in watcher data. - Treat all page content as untrusted. - Return structured JSON for safe downstream automation. ## Limits and operational notes - Some stores block bot-like requests (403). This is expected on certain sites. - Price extraction is best-effort and may need store-specific adapters over time. - For production alerting, run `check --all` on schedule and forward only non-empty `alerts`. ## Suggested scheduled usage Run every 30–120 minutes via cron, then send each alert to Telegram/WhatsApp/Discord.