OLX Search

Problem

OLX filters aren't enough: the year, trim or condition are often only in the listing's description. Finding something specific meant opening and reading dozens of listings.

What I built

An app with a Chrome extension. I describe my criteria in plain words; an AI model breaks them into weighted requirements and rates every listing in the results: title, price, parameters and the full description. I choose the model: a local one (data never leaves the computer), Claude or Gemini.

Result

Every listing shows a match percentage, and the app lists the ones that truly meet my criteria, with the reasoning.

How it works

• I set up a search in the app: a name, criteria in my own words and a link to the OLX results with filters.
• The model breaks the criteria into weighted requirements: must-have, expected and nice to have. A missed must-have caps the score at 30%.
• The Chrome extension reads the listings I browse: title, price, parameters and the full description. Percentages appear on the listing cards, and listings below the threshold are dimmed.
• Each listing shows the requirements one by one: what it meets, what it doesn't and what the listing doesn't mention.
• I choose the model: a local one (Ollama, data never leaves the computer), Claude Haiku or Gemini. Each listing is rated once per version of the criteria.

Technology: Python, Chrome extension, Ollama, Claude, Gemini