Skip to content

oniani1/E-Auction

Repository files navigation

E-Auction Deal Finder

Automated deal finder for eauction.ge — Georgia's government auction platform. Scrapes live auctions, analyzes them with AI (Gemini 2.0 Flash + vision), and surfaces the best deals.

What it does

  • Scrapes live auction listings from eauction.ge (search pages + full detail pages with all 4 tabs)
  • Extracts all property fields, descriptions, payment info, risks & conditions, images
  • Evaluates each listing with Gemini AI — deep analysis including condition assessment, red flags, market value estimation, and max bid recommendation
  • Surfaces deals ranked by score (1-10) in a dashboard
  • Notifies via Telegram when high-score deals appear (optional)

Tech Stack

  • Server: Node.js, TypeScript, Express, better-sqlite3 (WAL mode), Cheerio
  • Dashboard: React 18, TypeScript, Vite, Tailwind CSS v4, Zustand
  • AI: Google Gemini 2.0 Flash with vision (image analysis)
  • Notifications: Telegram Bot API (optional)
  • Monorepo: npm workspaces (server/ + dashboard/)

Setup

git clone https://github.com/nikaoniani/e-auction.git
cd e-auction
npm install

Create server/.env:

GEMINI_API_KEY=your_gemini_api_key

# Optional
TELEGRAM_BOT_TOKEN=your_telegram_bot_token
TELEGRAM_CHAT_ID=your_telegram_chat_id

PORT=3004
NODE_ENV=development
SCRAPE_INTERVAL_MINUTES=30
INITIAL_PAGES=10
SUBSEQUENT_PAGES=3

Running

# Development (server + dashboard)
npm run dev

# Server only
npm run dev:server

# Dashboard only
npm run dev:dash

# Production build
npm run build
npm start

How the AI Analysis Works

Each listing gets a deep evaluation covering:

  1. Item Identification — what exactly is being sold
  2. Property Details — area, cadastral code, infrastructure, utilities (for land/buildings)
  3. Condition Assessment — from image analysis + description parsing
  4. Red Flags & Risks — enforcement sale implications, hidden costs, legal risks, encumbrances
  5. Value Analysis — estimated market value based on Georgian market comparables
  6. Deal Verdict — score (1-10), max recommended bid, discount percentage

The scraper extracts data from all 4 tabs on each detail page:

  • Properties (address, cadastral code, area, utilities, assessment amount)
  • Description (full text, short description, conditions)
  • Payment information
  • Rules & conditions

Architecture

e-auction/
  server/
    src/
      ai/           # Gemini evaluation, prompts, types
      api/           # Express routes
      db/            # SQLite schema, migrations, queries
      notifications/ # Telegram bot
      scraper/       # Fetcher, search parser, detail parser
  dashboard/
    src/
      components/    # DealCard, DealList, Layout
      views/         # DealsView, AllListingsView, SettingsView
      stores/        # Zustand state
      lib/           # API client

License

MIT

About

Automated deal finder for eauction.ge — scrapes live government auctions, analyzes with Gemini AI + vision, and surfaces the best deals

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages