nepremicnine-monitor

A terminal-based real estate monitor for nepremicnine.net, Slovenia's largest property listing portal. Tracks listings across configurable searches, enriches them with public data, and estimates fair prices.

Features

TUI (Terminal User Interface)

Interactive Textual-based interface with three tabs:

• Listings — sortable table of all tracked listings showing price, size, year, distance to Ljubljana, estimated fair price, GURS valuation, description tags, and more
• Summary — aggregated view grouped by region, municipality, and property type with averages and counts (deduplicates cross-listed properties)
• Log — live scraping progress

Key bindings:

Key	Action
f	Open filter dialog (year, price, size, distance, categories)
r	Trigger immediate scrape of all monitors
1/2/3	Switch tabs
Enter	Open listing in browser
q	Quit

Scraping

• Uses curl_cffi with Chrome impersonation to bypass Cloudflare
• Paginated search result scraping with duplicate page detection
• Detail page scraping for full listing data (location, description, year adapted)
• Configurable crawl delay and request timeout
• Automatic periodic re-scraping at a configurable interval

Data Enrichment

GURS Cadastral Data

Queries the GURS (Geodetska uprava RS) cadastral API to match listings with official building records:

• Matches by settlement name + building area (m2) within a configurable tolerance
• Retrieves official building valuation (posplosena vrednost) for price comparison
• Handles Ljubljana district normalization (e.g. "LJ-Vic" to "Vic")
• Falls back to sub-area extraction from listing titles when settlements exceed the API's 200-result limit
• Match statuses: matched, multiple_matches, no_match, too_many_results, no_settlement

Geocoding & Distance

• Geocodes municipalities via Nominatim (OpenStreetMap)
• Computes driving distance to Ljubljana via OSRM, with Haversine fallback
• Results cached in SQLite to avoid repeated API calls

Description Tagging

Regex-based extraction of structured tags from Slovenian listing descriptions:

• Permits: use_permit, no_use_permit, building_permit, no_building_permit
• Legal: illegal_build, under_construction
• Condition: needs_renovation, renovated
• Financial: investment, rental_tenants
• Amenities: garage, parking
• Energy: energy_class_A1 through energy_class_G

Negative signals override positive ones (e.g. "nima uporabnega dovoljenja" suppresses use_permit).

Price Estimation

OLS (Ordinary Least Squares) regression trained per property type in log-price space:

• Houses: features are building size, land size, effective age, driving distance
• Apartments: features are apartment size, effective age, driving distance, floor number
• Outlier filtering (2.5 sigma on price/m2) before training
• Uses year_adapted over year_built for effective age when available
• Minimum 8 listings per group to train; no external dependencies (pure Python solver)

Installation

Requires Python 3.11+.

pip install -e .

For development (tests, linting):

pip install -e ".[dev]"

Configuration

Create a config.toml file:

[[monitors]]
name = "Apartments Slovenia"
url = "https://www.nepremicnine.net/oglasi-prodaja/slovenija/stanovanje/?s=16"
max_pages = 25

[[monitors]]
name = "Houses Slovenia"
url = "https://www.nepremicnine.net/oglasi-prodaja/slovenija/hise/?s=16"
max_pages = 25

[scraper]
crawl_delay_seconds = 5
request_timeout_seconds = 30

[database]
path = "nepremicnine.db"

[monitor]
interval_minutes = 60

Each [[monitors]] entry defines a search URL to track. The url should be a nepremicnine.net search results page; max_pages controls how deep pagination goes.

Usage

TUI mode (default)

nepremicnine-monitor config.toml

CLI commands

# Backfill listings from a specific search URL
nepremicnine-monitor config.toml backfill-search "https://www.nepremicnine.net/oglasi-prodaja/..."

# Geocode all municipalities and compute driving distances
nepremicnine-monitor config.toml backfill-distances

# Enrich listings with GURS cadastral data (optional tolerance in m2)
nepremicnine-monitor config.toml enrich-gurs 2.0

# Extract description tags for all untagged listings
nepremicnine-monitor config.toml tag-descriptions

Project Structure

nepremicnine_monitor/
  __main__.py          CLI entry point, subcommand dispatch
  scraper.py           Async HTTP client with Cloudflare bypass
  parser.py            HTML parsing for search results and detail pages
  db.py                SQLite database layer (thread-safe, WAL mode)
  tui.py               Textual TUI application
  geocoder.py          Nominatim + OSRM geocoding and distance
  gurs_enricher.py     GURS cadastral API integration
  description_tagger.py  Regex-based tag extraction from Slovenian text
  pricing.py           OLS fair price estimator
tests/
  test_db.py           Database CRUD and query tests
  test_parser.py       HTML parsing tests
  test_scraper.py      Pagination and dedup tests
  test_description_tagger.py  Tag extraction tests
  test_pricing.py      OLS estimator tests

Development

# Run tests
pytest tests/ -v

# Format
black nepremicnine_monitor/ tests/
isort nepremicnine_monitor/ tests/

# Lint
pylint nepremicnine_monitor/

Pre-commit hooks are configured for black, isort, and pylint.

License

Private project.