[04]

Gel Gez Gor

2026

Client

Gel Gez Gör

Year

2026

Scope

Automation, WhatsApp, social media content

Role

Automation / Development

§ Case Study

Gel Gez Gor — Trend detection, WhatsApp notifications and social content automation for the Gel Gez Gor classifieds site.

We built an automation layer for Gel Gez Gor, a classifieds site in North Cyprus, that detects trending listings automatically. The platform itself belongs to the client; our work was the pipeline that reads featured listings from the site and notifies listing owners via WhatsApp.

The same system generates a ready-to-post social media visual and an AI-assisted caption for each listing and can publish them to Instagram and Facebook. The team now runs the monitoring and posting work it used to do by hand from a single panel.

AutomationWhatsAppSocial MediaAI

Challenge

At Gel Gez Gor, manually tracking trending listings, reaching out to listing owners one by one, and preparing a separate social media graphic for every listing took time and produced inconsistent results. The platform itself had no API for trend detection. What was needed was an automation layer that could detect trending listings straight from the site's own HTML structure, notify listing owners over WhatsApp, and generate a ready-made Instagram/Facebook graphic and caption for each listing — one that wouldn't break easily if the site's structure changed, and that would run on its own several times a day.

Solution

The pipeline reads gelgezgor.com's static HTML with Cheerio to extract trending listing cards; results are saved locally to SQLite/Drizzle, or directly to a cloud PostgreSQL database in the Vercel environment, then enriched from the detail page. Listing owners get automatic notifications via the Twilio WhatsApp Business API, with a separate client already in place for migrating to Meta's own WhatsApp Cloud API. On the social side, a Sharp-based SVG-overlay-and-image-composite approach produces six graphic templates across three themes (Modern, Elegant, Bold) times two variants (dark/light); template choice auto-rotates based on the listing's title and category. Captions and hashtags are generated first with Gemini, with an automatic fallback to template-based text if the API call fails; an optional AI-generated background via Imagen 3 is also attempted, again falling back to the template. Generated graphics can be published directly to Instagram (single post and carousel) and Facebook through the Meta Graph API. All of this comes together in a single orchestrator triggered twice a day by node-cron; an Express-based panel provides manual triggering and status tracking, with a dry-run mode for testing without touching production.

Key technical decisions

A Cheerio-based scraper with documented selectors reads trending listing cards straight from the site's static HTML — no headless browser required.
Automatic listing-owner notifications via the Twilio WhatsApp Business API, with a separate client already in place for migrating to the Meta WhatsApp Cloud API.
A Sharp-based graphics engine auto-rotates between six templates (3 themes × 2 variants) based on listing category.
Caption and hashtag generation with Gemini, backed by an automatic fallback chain to template-based text on failure.
Generated graphics publish directly to Instagram (single post and carousel) and Facebook through the Meta Graph API, with a dry-run mode for safe testing before going live.

Tech stack

TypeScriptNode.jsExpressCheerioDrizzle ORMSQLitePostgreSQLTwilioSharpGoogle Gemini

Outcome

The result is a pipeline that unifies trend detection for Gel Gez Gor into a single automated flow, from scraping through WhatsApp notification to social media publishing. Manual control and a dry-run mode in the panel allow safe testing before going live. The system was built to be portable, running equally well on SQLite on a single machine or on cloud PostgreSQL on Vercel.