Radar de Precios

Problem

The same medicine can cost 20-70% more at one Peruvian pharmacy chain than another, and two of the country's three largest chains belong to the same holding company — nothing forces price transparency, and no public tool lets a buyer compare instantly before they walk into a store.

Approach

Per-chain adapters over undocumented APIs (Algolia, SFCC) and a three-layer matcher that decides, without a shared product catalog, when two listings are the same medicine.

Inkafarma and Mifarma belong to the same holding and run on the holding’s own Algolia search, so their listings share an internal product ID and match for free. Boticas Perú runs on Salesforce Commerce Cloud instead, with no shared ID at all — its adapter combines the public HTML product grid with the storefront’s own QuickView JSON. None of these are official APIs: each adapter reads the same public, search-only client keys and endpoints the chains’ own storefronts call from the browser, throttled well below anything that would look like abuse.

Matching across chains with no shared ID is the actual problem the project solves. The matcher runs in three layers: exact ID where it exists, fuzzy matching on name, active ingredient and pack size (guarded against bundles and near-miss variants like different vitamin doses) where it doesn’t, and perceptual image hashing as a final check on doubtful fuzzy matches. Every run is kept as its own snapshot rather than overwritten, so price history and promo detection are possible later without re-scraping the past.

The published site is deliberately boring: static HTML/CSS/JS with no build step, reading a JSON file at runtime and rendering the comparison table client-side, so it costs nothing to host and drops into any subdirectory unchanged.