FlipFinder — AI search for profitable iPhones on OLX

Resellers scroll OLX by hand, keep prices in their heads and miss the best deals: a profitable iPhone is gone within minutes. The goals were to:

• continuously collect new listings without bypassing the marketplace's protection;
• know the real market price of a specific model, storage and condition instead of guessing;
• filter out suspicious listings: someone else's iCloud, model mismatch, odd descriptions;
• send only real resale candidates, with no spam.

We built a seven-step pipeline:

1. Collection through an ordinary local browser (Playwright + Chrome), with no CAPTCHA bypass, proxies or cookie theft.
2. Normalisation: model, storage, condition, battery and accessories are extracted from the listing text.
3. Market base over 72 hours with benchmarks by model, storage and condition.
4. Pre-ranking — the expensive AI analysis runs only for the best candidates.
5. AI analysis (Groq or any OpenAI-compatible model): risk signals, pros, inconsistencies in the description.
6. Deterministic estimate of resale price, profit and margin, kept separate from AI advice.
7. Telegram reports and a local web panel: market overview, favourites, notes, comparison, price history and trends, price drops.

The interface and notifications are in Russian and Polish, with separate language settings for the panel and Telegram. The project is published on GitHub under AGPL-3.0, covered by tests and runs on Windows in five commands.