AI Car Recommendation Agent — Multi-Marketplace Search

Overview

An AI-powered virtual sales advisor built over the three biggest Saudi car marketplaces — Dubizzle.sa, Syarah and Haraj. A scraping pipeline runs on a Linux server at regular intervals to keep roughly 10,000 active listings — and their images — current in a Supabase database, and the agent searches that index rather than the live sites. Buyers reach it through a chat interface and talk to it the way they would to a salesperson; it understands them in English and Arabic, filters across price, brand, mileage and condition, and surfaces ranked alternatives when there is no exact match. Listings and their photos are also browsable on the React comparison platform built alongside it.

Outcome

Roughly 10,000 validated listings from Saudi Arabia's three largest car marketplaces kept current on a scheduled refresh — searchable by chatting with a sales agent in English or Arabic, and browsable with photos on the React comparison platform, in place of checking three sites by hand.

Screens & Demos

AI Car Recommendation Agent — Multi-Marketplace Search
AI Car Recommendation Agent — Multi-Marketplace Search

Architecture & Pipeline

  1. 01

    Scheduler

    Linux server · regular intervals

    • Linux
  2. 02

    Dubizzle.sa · Syarah · Haraj

    Source marketplaces

  3. 03

    Scrapers

    Python · Selenium · BeautifulSoup

    • BeautifulSoup
    • Python
    • Selenium
  4. 04

    Validation Filters

    Price · condition · availability

  5. 05

    Supabase / PostgreSQL

    Listings + images · ~10k active

    • PostgreSQL
  6. 06

    Chat Interface (EN/AR)

    Conversational sales agent

  7. 07

    Intent Parser

    OpenAI GPT-4 · LangChain

    • GPT-4
    • LangChain
    • OpenAI
  8. 08

    Filter Builder

    Price · Brand · Mileage · Condition

  9. 09

    Ranker

    Scoring + fallbacks

  10. 10

    React Platform / API

    FastAPI · Docker · comparison UI

    • Docker
    • FastAPI
    • React

End-to-end flow, in order. Scroll sideways on a narrow screen.

Key Features

  • Chat interface where a buyer converses with the agent like a salesperson
  • Natural-language query understanding with filtering across multiple parameters
  • Ranked recommendations with fallback alternatives when nothing matches exactly
  • Continuous scraping of Dubizzle.sa, Syarah and Haraj — roughly 10k active listings
  • Validation filters for price, condition and availability before a listing is indexed
  • Image extraction tied to each structured listing record
  • Scheduled refresh on a Linux server keeps listings and prices current
  • Supabase / PostgreSQL for the listing database and image storage
  • React comparison platform browsing the indexed listings with their photos
  • Bilingual support (English and Arabic)
  • Designed to scale across additional marketplaces