Local PC/Raspberry Pi platform that aggregates real estate listings (sale and rental) from Immobiliare.it and Idealista, deduplicates multiple ads for the same property, filters out unwanted listings (e.g., bare ownership, ground floors, or court auctions), and sends real-time notifications on Telegram and/or Email.
Real estate portals erase history and hide metrics to protect listing agencies. This platform stores everything in a local SQLite database (case.db), acting as your personal buyer decision engine and negotiation assistant:
- Deal Radar & Congruity Engine (
Deal Score): Automatically compares every new listing against the localized €/sqm median of its specific micro-zone (pricing_stats.py). Combined with agency historical discounting behavior (market_velocity.py), it reveals undervalued opportunities within minutes of publication. - Ghost Price & Re-listing Memory (
Recycled Ad Tracker): When a house sits unsold for months at €420k, agencies often delete the ad and re-publish it weeks later as "New" at €389k with different photos. Our cross-portal deduplicator matches coordinates and sqm across oldgonelistings to alert you:[RECYCLED PROPERTY] previously listed for 160 days at higher price (-9.5%). - Red Flags Audit & Real TCO Calculator: Extracts structural and legal red flags directly from ad text (e.g., active tenant lease, bare ownership, high condo fees > €250/mo, no elevator) and calculates your realistic monthly Total Cost of Ownership (Mortgage + Condo Fees + Renovation buffer).
- Zero Cloud, 100% Local Privacy: Runs locally on your own PC or Raspberry Pi. No paid APIs, no subscriptions, no tracking.
Double-click on scripts\windows\start.bat:
- Installs all dependencies on first run (requires Python 3.11+ and Node.js 18+).
- Starts the backend server (http://localhost:8000) and the frontend dashboard (http://localhost:5173).
- Automatically opens the web interface in your default browser.
All Windows-only helpers (service install/uninstall, restart, stop, hidden
autostart) live in scripts\windows\ — see Remote Access & Running in the
Background.
Open a terminal inside the project directory and run:
chmod +x scripts/linux/start.sh
./scripts/linux/start.sh- Installs dependencies and starts both services concurrently.
- Makes the dashboard accessible from any device on your local network at
http://<IP_OF_YOUR_PI>:5173.
The scraper stays on the PC — portals trust residential IPs and block cloud ones
— but the dashboard works from an Android or iOS browser, installable as an app
icon. Run scripts\windows\serve.bat instead of start.bat: it builds the
frontend and serves the dashboard and API from a single port (8000), so there
is nothing to configure on the phone. Reaching it from outside the house, the
security model of the open (unauthenticated) API, and the optional API token
are covered in Remote Access & Running in the Background.
- Get the search URL from Immobiliare.it or Idealista.it (or describe the search in plain Italian and let the built-in assistant build both URLs for you), and copy it from the address bar.
- Add Profile: paste the URL into "+ Add search profile", name it, save.
- Start Scanning: click "Start Scan Now", or let the scheduler run in the background.
- Browse Listings: merged duplicates show a purple badge, and properties found since your last visit carry a 🆕 new badge.
- Curate: hide listings you don't want, mark ones as sold/rented, search and filter the grid, and clean up in bulk.
The full walkthrough — every portal filter, search-building shortcuts, the search/filter bar, bulk cleanup, deleting a search (and what happens to the listings it found), and silencing a search's notifications — is in Using the App.
The dashboard has a full toolbox beyond the grid itself: a bilingual interface, a map view with drawable filter zones, price-fairness and Deal Score checks, a Smart Match Score against your "dream home", price-trend and market-velocity charts, a scraper-health panel, free-form tags, shortlist exports (HTML/MD/CSV), a mortgage calculator, and an email inbox import that can pull in listings from your own mailbox. See Features for the full rundown, and Email Inbox Import for that last one specifically.
- Data Persistence: All settings, search profiles, listings, price history, and hidden statuses are saved locally in a database file (
backend/case.db). You can close the app or shut down your PC at any time without losing any progress or history. - Always-on Scanning: background scans, price-trend snapshots, and alerts run only while the app is running. With
start.bat/serve.batthat means keeping the terminal window open (minimized). To run it silently with no window, see Running it 24/7 on Windows. - Restart from the dashboard: after updating the app, press Settings → 🔄 Restart backend instead of hunting for the terminal window. The dashboard goes offline for a few seconds and reloads itself. With
start.bat(which runs with auto-reload) a code update is usually picked up on its own; withserve.batthis button is the quick way to apply one. (It restarts the backend only — withserve.bat, which serves a pre-built frontend, re-run the script if the frontend itself changed.) - Pause automatic scans: Settings → Automatic scan → Pause automatic scans stops the scheduled scans from touching the portals — handy for resting the connection while you are away, without deactivating every search one by one. The top bar shows ⏸ Automatic scans paused while it is on, and Start Scan Now still works on demand (an explicit request bypasses the pause). To silence a single search instead, untick it in the search list.
- Catch-up Scan: the scheduled scan normally fires one full interval after startup. If the PC was off and the last scan is already older than the configured interval, a catch-up scan runs ~2 minutes after startup instead — so switching the PC on is enough to bring the listings up to date.
- Automatic Backups: a copy of
case.dbis written tobackend/backups/at most once per day (checked at startup; the 14 most recent copies are kept). The folder is local — point your cloud-sync or a second drive at it if you want off-machine safety. - In-App Log Viewer: click 📜 in the top bar to see the backend's own log — scan progress, the availability check advancing line by line, DataDome blocks — without opening
backend/app.login a text editor. It filters by keyword and auto-refreshes every few seconds while open, so you can tell a slow-but-working check apart from a genuinely stuck one. - Data Management (start fresh): Settings → Data management has four irreversible resets, each behind a confirmation. Reset email imports clears everything the inbox import staged, so you can import again from scratch. Clear dashboard deletes all found properties and price history but keeps your search profiles — the next scan rebuilds the grid silently (no notification flood). Clear price trends drops only the trend-chart history. Factory reset wipes everything back to a fresh install (a backup of the database is saved first). Your notification and login settings are never touched.
Telegram and Email are both configured in Settings, with a step-by-step guide next to each, and each search profile can route its own alerts to either channel, both, or neither. See Notifications for the setup details, using Gmail's app password, and scraper-health alerts.
The first scan of any search profile retrieves all existing listings and saves them to build the baseline database. No notifications are sent during the first scan to avoid flooding your Telegram account. You will only receive alerts for new listings and price drops starting from the second scan onward.
Both portals sit behind DataDome, an anti-bot system that occasionally
blocks a scan instead of returning listings — expected, not a bug: a blocked
profile is retried on the next scan, and you're alerted only after several
failures in a row. The most effective fix is handing the scraper a real
browser's datadome cookie, which the app can grab for you automatically; a
proxy pool or a paid scraping API are further options for a stubborn block. See
Keeping Scans Unblocked for all the levers, from most to
least automatic.
- Backend: Python 3.11+ / FastAPI / SQLite / APScheduler.
- Resilient Scrapers: Built on 4 fallback strategies (JSON-LD Schema → Embedded
__NEXT_DATA__state → Heuristic class-free HTML parsing → Internal API fallback). - Residential IP Scraping: Designed to run locally or on home networks. Cloud server IPs are heavily blocked by DataDome, whereas your home internet IP is trusted, ensuring reliable scans.
- Deduplication Engine: Listings are merged only if they contain geographical proof (coordinates within 60 meters OR exact same street and house number) plus compatible price, rooms, floor, and square meters.
- Frontend: React / Vite / TypeScript / Tailwind CSS, bilingual (English / Italian) through a small dependency-free dictionary — no i18n library, and a key present in one language but missing in the other fails the build.
Automated tests cover all parser strategies, price formatting edge cases, deduplication rules, price history changes, and scanner routines — all offline (no network calls), so they always pass or fail for a real reason. The frontend has its own unit tests for the pure logic (filter querystring codec, floor labels, and the English/Italian dictionaries — key and placeholder parity).
Run the backend tests using the local Python virtual environment:
cd backend
& .venv/Scripts/python.exe -m pytestRun the frontend tests:
cd frontend
npm test(All tests must pass before committing changes).
Beyond the runtime dependencies, an optional dev toolchain (linting, coverage,
property-based tests, dependency CVE scanning, and a pre-commit hook) lives in
backend/requirements-dev.txt. It is never installed on the target device —
only in a development checkout:
cd backend
& .venv/Scripts/python.exe -m pip install -r requirements-dev.txt
& .venv/Scripts/ruff.exe check app tests # lint
& .venv/Scripts/ruff.exe format app tests # format
& .venv/Scripts/python.exe -m pip_audit -r requirements.txt # CVE scan