Tampa MLS
The luxury end of the same brokerage: Hillsborough, Pinellas and Pasco inventory with a $400,000 floor written into the search layer itself.
Overview
Tampa MLS is the luxury end of a Tampa Bay brokerage's property search: Hillsborough, Pinellas and Pasco inventory, filtered search, listing pages with photos, maps and school data, and a member area where buyers save homes and searches and send questions straight to the agent team. The luxury focus is enforced in the search layer itself - unless a visitor sets a minimum price, results start at $400,000, and the homepage leads with price bands from $600,000 up.
It runs as a custom PHP application on Slim 3 with Smarty templates. Listings live in MongoDB as documents keyed by MLS number; members, favorites, saved searches, contact requests and editable page metadata live in MySQL; sessions sit in Redis; outgoing email goes through a Beanstalkd job queue. The site is served through Cloudflare and still answers on its original PHP runtime.
Around the search sit the pages a brokerage needs: area information, selling, FAQ, about and contact - the same application rendering content pages from the same templates.
What was built
- A search layer that turns 16 filter types into database queries: price, beds, baths, square footage, lot size, year built, garage, property type, pool type, sale type, county, city, zip code, school, golf course and 19 kinds of waterfront frontage, with 16 sort orders.
- Path-based search URLs such as price, beds or zip codes as URL segments, so every filtered result page can be bookmarked, shared and paginated without losing its filters.
- Ready-made searches for waterfront, beachfront, golf and new homes, plus homepage shortcuts by price band and by urban area.
- Listing pages with a photo gallery, property, interior and exterior features, assigned schools, a hybrid Google map and a Street View panorama that is checked for availability before it is shown.
- A member area with registration, login with a remember-me cookie, password recovery, favorites, recently viewed homes and saved searches with instant, daily or no email alerts.
- Lead capture on every listing card and page (request more info, schedule a showing) and a contact form protected by Google reCAPTCHA, with staff notifications sent through the email queue.
- Open Graph place and house tags on listing pages carrying address, coordinates, beds, baths and square feet, so a shared link shows the home rather than a bare title.
- Maintenance jobs that generate the XML sitemap, rebuild the school directory used by the school filter from listing data, and bundle plugins, minified app code and client-side templates into one JavaScript file.
Stack
Frontend
- Smarty 3
- Server-rendered templates for search, listing, member and content pages
- Semantic UI
- Cards, modals, dropdown filters and forms
- jQuery + JsRender
- Search form logic, favorites, saved-search editing and client-side modal templates
- Fotorama
- Swipeable photo gallery, loaded only on listing pages to keep search pages light
Backend
- PHP / Slim 3
- Routing, middleware and dependency container for pages, search, member area and AJAX endpoints
- Beanstalkd (Pheanstalk)
- Email jobs: the message is stored first, then its id goes on a delivery queue
Data
- MongoDB
- Listing documents keyed by MLS number; filters, sorting and pagination run as queries
- MySQL
- Members, favorites, saved searches, contact requests and per-page SEO metadata
- Redis
- Session storage for signed-in members, including remember-me logins across visits
Infrastructure & DevOps
- Cloudflare
- The site is served through Cloudflare
Integrations
- MLS listing data
- Property facts, photos, schools and listing office behind every search card and listing page
- Google Maps + Street View
- Hybrid map and street-level panorama for each listing address
- Google reCAPTCHA
- Spam protection on the contact form, verified on the server
Engineering highlights
Default business rules sit in one place in the search layer: only active listings, the three core counties unless others are chosen, and a $400,000 floor unless the visitor sets a minimum price.
All search state lives in the URL path, which is what lets a member save any result page as a saved search and lets the homepage link straight to curated searches.
Member data is partitioned by a site identifier, so one member database can back more than one property site - which is what made the second site possible.
The document store fits the domain: a listing is one MLS-keyed document with its photos, features and schools, so a search is a query rather than a join across a dozen tables.