Telegram bot for auto auctions 24/7 on a Korean car website
Bot Development
Client Pain Point
Good lots in the car market disappear within hours. The client purchased cars on the Korean marketplace Encar: over 200 listings per day, all in Korean, with the main risks—accidents, flooding, rental history—hidden in separate inspection and insurance history reports. Manual monitoring meant either sleepless nights or lost best options.
What Was Done
A Telegram bot that queries Encar every 30–60 minutes around the clock
Reverse engineering of 4 private API endpoints (search, lot details, technical inspection, insurance history)—without a browser and Selenium, quickly and reliably
A filter with 10 criteria: model, configuration, year, mileage, price, fuel + automatic checking for red flags: accidents, flooding, rental/lease
Each match arrives as a card in Telegram: all photos + structured summary in Ukrainian with history badges
SQLite deduplication—no lot is sent twice; failed requests are retried in the next pass
Result
The client reviews 1–2 filtered cars per day instead of 200+ listings
Cars with hidden accidents/flooding/rentals are automatically filtered out
Saved ~2 hours of manual monitoring daily, no lost lots since launch
Operates unattended on VPS as a systemd service
Stack
Python 3.12 · httpx (async) · aiogram 3 · SQLite · pydantic-settings · structlog · systemd/VPS
Good lots in the car market disappear within hours. The client purchased cars on the Korean marketplace Encar: over 200 listings per day, all in Korean, with the main risks—accidents, flooding, rental history—hidden in separate inspection and insurance history reports. Manual monitoring meant either sleepless nights or lost best options.
What Was Done
A Telegram bot that queries Encar every 30–60 minutes around the clock
Reverse engineering of 4 private API endpoints (search, lot details, technical inspection, insurance history)—without a browser and Selenium, quickly and reliably
A filter with 10 criteria: model, configuration, year, mileage, price, fuel + automatic checking for red flags: accidents, flooding, rental/lease
Each match arrives as a card in Telegram: all photos + structured summary in Ukrainian with history badges
SQLite deduplication—no lot is sent twice; failed requests are retried in the next pass
Result
The client reviews 1–2 filtered cars per day instead of 200+ listings
Cars with hidden accidents/flooding/rentals are automatically filtered out
Saved ~2 hours of manual monitoring daily, no lost lots since launch
Operates unattended on VPS as a systemd service
Stack
Python 3.12 · httpx (async) · aiogram 3 · SQLite · pydantic-settings · structlog · systemd/VPS