MietRadar 📡

Automated Apartment & WG Application Bot for Germany — WG-Gesucht + ImmobilienScout24

Scan. Apply. Move in. — An AI-powered Python bot that monitors WG-Gesucht and ImmobilienScout24, automatically applies to new rental listings, and personalises each application using an LLM — either fully local, or via a hosted API (Google Gemini, OpenAI).

---

What is MietRadar?

Finding an apartment in Germany — especially in cities like Munich, Berlin, or Hamburg — is brutally competitive. Listings disappear within minutes. MietRadar monitors WG-Gesucht and ImmobilienScout24 around the clock, instantly applies to new listings matching your search filters, and optionally uses an LLM to write a personalised paragraph for each application. Choose the backend that fits you: a local Gemma 3 4B-IT model (no API costs, runs on your machine), or a hosted API — Google Gemini (Developer API key or Vertex AI) or OpenAI (e.g. GPT-5 nano).

Keywords: Wohnung Bot, WG Bot, apartment bot Germany, Wohnungssuche automatisieren, WG-Gesucht Bot, ImmobilienScout24 Bot, rental application automation, Mietwohnung Bot, automated apartment search Germany, Bewerbung automatisch senden

Disclaimer: Use at your own risk. Respect the platforms' terms of service. This tool is provided "as is" without warranty.

Features

Feature	Description
🏠 Dual Platform	Supports both WG-Gesucht and ImmobilienScout24
🔄 Zero-Touch Loop	Scans search results every N minutes, applies to new listings automatically
🥷 Stealth Automation	Chrome with anti-detection patches, human-like typing & clicking
🚫 Smart Blacklisting	Auto-blocks sent listings, skips sponsored/agency ads, manual override lists
🧠 AI Personalisation	Multi-provider LLM (local Gemma 3 4B-IT, Google Gemini, or OpenAI) generates a unique paragraph per listing (optional)
📊 Reply Tracking	CSV reports with 🟢/🟡/🔴 status for each application
⚡ Shared Pipeline	Single message template, LLM persona, and helper functions for both platforms

Directory Structure

miet-radar/
├── config/                         # User-editable configurations
│   ├── .env                        # Secrets (credentials, URLs, HF token)
│   ├── .env.example                # Template for .env
│   ├── llm_persona.txt             # LLM prompt/identity (editable)
│   ├── message.txt                 # Shared message template (both bots)
│   ├── message.txt.example         # Template for message.txt
│   ├── wg_blacklist.txt            # Manual blacklist for WG-Gesucht
│   ├── wg_blacklist.txt.example    # Template
│   ├── immo_blacklist.txt          # Manual blacklist for ImmoScout24
│   └── immo_blacklist.txt.example  # Template
├── data/                           # Runtime data (auto-generated, DO NOT EDIT)
│   ├── wg_diff.dat                 # Processed WG IDs
│   ├── wg_sent_request.dat         # WG audit log with timestamps
│   ├── wg_replies_report.csv       # WG reply tracker
│   ├── wg_offer.json               # WG scan snapshot
│   ├── wgbot_profile/              # WG browser session cache
│   ├── immo_diff.dat               # Processed ImmoScout IDs
│   ├── immo_sent_request.dat       # ImmoScout audit log
│   ├── immo_replies_report.csv     # ImmoScout reply tracker
│   ├── immo_offer.json             # ImmoScout scan snapshot
│   └── immobot_profile/            # ImmoScout browser session cache
├── src/                            # Core Python application logic
│   ├── wg-gesucht.py               # Main loop — WG-Gesucht
│   ├── submit_wg.py                # WG browser automation & form submission
│   ├── check_replies.py            # WG reply tracker
│   ├── immoscout.py                # Main loop — ImmobilienScout24
│   ├── submit_immo.py              # IS24 browser automation & form submission
│   ├── check_replies_immo.py       # IS24 reply tracker
│   └── llm_personalizer.py         # Shared LLM inference (Gemma 3 4B-IT, Gemini, OpenAI)
├── tests/                          # Tests (no live sends)
│   ├── test_llm_personalizer.py    # WG end-to-end test
│   ├── test_immo.py                # IS24 end-to-end test
│   ├── test_stealth.py             # Browser stealth validation
│   └── test_llm_only.py            # Isolated LLM test
├── scripts/
│   └── setup.sh                    # One-command setup script
├── README.md
└── requirements.txt

Quick Start

1. Prerequisites

• Python 3.10+
• Google Chrome installed

2. Setup (One Command)

git clone https://github.com/unik-w/MietRadar
cd miet-radar
bash scripts/setup.sh

The setup script creates a virtual environment, installs all dependencies, and scaffolds config files from templates.

2b. Manual Install (Alternative)

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp config/.env.example config/.env
cp config/message.txt.example config/message.txt

3. Configuration

1. Credentials — Edit config/.env:

   • WG-Gesucht: WG_EMAIL, WG_PASSWORD, WG_SEARCH_URLS
   • ImmoScout24: IMMO_EMAIL, IMMO_PASSWORD, IMMO_SEARCH_URLS and personal details (IMMO_FIRST_NAME, IMMO_LAST_NAME, etc.) for the IS24 contact form.
   • (Optional) HF_TOKEN for LLM Personalisation.

2. Message Template — Edit config/message.txt:

   • Use {name} for the landlord's name (auto-filled)
   • Use {LLM_TEXT} for the AI-generated paragraph (auto-filled when LLM is enabled)

3. LLM Persona (Optional) — Edit config/llm_persona.txt: Set USE_LLM_PERSONALIZATION=true in .env to enable. The persona auto-adapts to WG vs. apartment listings.

   Choose the LLM backend via LLM_PROVIDER in .env:

   LLM_PROVIDER	Backend	Required .env vars
   gemma_local (default)	Local Gemma 3 4B-IT (HuggingFace + PyTorch, runs on your machine)	HF_TOKEN
   gemini	Google Gemini API — Developer API key or Vertex AI	GEMINI_MODEL + either GEMINI_API_KEY or (GOOGLE_GENAI_USE_VERTEXAI=true, GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_LOCATION, and gcloud auth application-default login)
   openai	OpenAI API (e.g. GPT-5 nano)	OPENAI_MODEL, OPENAI_API_KEY

   Install the extra SDKs only for the backend(s) you use: pip install google-genai (Gemini) and/or pip install openai (OpenAI).

   💸 Free credits tip: new Google Cloud accounts get $300 in free credit (valid ~90 days), usable against Vertex AI usage — not just Gemini. Vertex AI's Model Garden also hosts third-party models (e.g. Anthropic Claude) and open-source models (Llama, Mistral, and others), so the same credit can be used to experiment with those too, not only Google's own models. To use Vertex AI instead of a plain Gemini API key:

   1. Install the gcloud CLI: brew install --cask google-cloud-sdk (macOS)
   2. gcloud auth application-default login (one-time browser login)
   3. gcloud services enable aiplatform.googleapis.com --project=<your-project-id>
   4. Set GOOGLE_GENAI_USE_VERTEXAI=true, GOOGLE_CLOUD_PROJECT, and GOOGLE_CLOUD_LOCATION in .env

   Note: model availability varies by project/region on Vertex AI — if a model 404s, check the Model Garden for exact IDs enabled for your project, or try location=global for newer preview models.

4. Blacklists:

   • config/wg_blacklist.txt — paste WG-Gesucht URLs to ignore
   • config/immo_blacklist.txt — paste ImmoScout expose IDs to ignore

4. First Run & CAPTCHAs 🧩

On first run, the platform will likely show a CAPTCHA. The bot detects this and pauses — solve it manually in the browser window. Subsequent runs use saved session cookies from data/wgbot_profile/ or data/immobot_profile/.

Usage

Run the WG-Gesucht Bot

source venv/bin/activate
python src/wg-gesucht.py

Run the ImmobilienScout24 Bot

source venv/bin/activate
python src/immoscout.py

Both bots run independently as continuous loops. They scan multiple search result pages, apply to new listings, update blacklists, and sleep for CHECK_INTERVAL_SECONDS (configurable in .env).

Reply Tracking 📊

Generate a report of all sent applications and their reply status:

Command	Report File
python src/check_replies.py	data/wg_replies_report.csv
python src/check_replies_immo.py	data/immo_replies_report.csv

Status indicators:

• 🟢 Replied — host has responded
• 🟡 Pending — no reply yet, sent < 3 days ago
• 🔴 No Reply — no response after 3+ days

Testing (Dry Run)

Test the full pipeline without sending any messages:

# WG-Gesucht — scrape + LLM personalisation
python tests/test_llm_personalizer.py --listings 3

# ImmobilienScout24 — scrape + LLM personalisation
python tests/test_immo.py --listings 3

# ImmoScout24 — template only, no LLM (fast scraping test)
python tests/test_immo.py --no-llm --listings 3

# Browser stealth validation
python tests/test_stealth.py

---

How It Works

┌─────────────┐    ┌──────────────┐    ┌───────────────┐    ┌──────────────┐
│  Scan search │───▶│ Filter: skip │───▶│ Extract desc  │───▶│ Generate msg │
│  result pages│    │ sponsored &  │    │ + contact name│    │ (template +  │
│  (paginated) │    │ blacklisted  │    │ from expose   │    │  LLM paragraph│
└─────────────┘    └──────────────┘    └───────────────┘    └──────┬───────┘
                                                                   │
                                                                   ▼
                                                           ┌──────────────┐
                                                           │ Fill form &  │
                                                           │ send message │
                                                           │ (stealth)    │
                                                           └──────────────┘

---

Credits & History 📜

This project is an advanced evolution of the original immo repository by nickirk.

While MietRadar has been completely restructured and modernized with Python 3 support, dual-platform automation, Selenium stealth techniques, and LLM-based personalization, the original git history has been merged and preserved to ensure proper attribution to the foundational work of the original contributors.