Local business lead discovery & enrichment. Pick a city and a set of sectors LeadRadar finds businesses with missing or broken websites, scores the opportunity from verifiable evidence, and produces a sales-ready PDF report.
Built for web designers, digital agencies, and local B2B sales teams.
🇹🇷 Türkçe sürüm: README.tr.md
Pick a city, optional districts and any number of sectors then start the scan and watch progress stream live in the browser.
- 🗺️ Discovery Scans businesses across 27 German cities and 25 sectors via OpenStreetMap (Overpass API). No API key required, completely free.
- 🔍 Website audit Tests each prospect's site: HTTPS, mobile-friendliness (viewport), working contact paths, broken booking links, page errors.
- 🎯 Evidence-based scoring HIGH / MEDIUM / LOW opportunity score. Scores are derived only from verified findings; subjective impressions ("looks outdated") are tracked separately and can never on their own produce a HIGH score.
- 🧠 Deduplication Previously reported businesses never appear again (SQLite fingerprint memory).
- 📄 PDF report One page per prospect: parameters → contact details and an outreach plan → website issues with a technical audit table.
- 🖥️ Local control panel Pick city, districts, and sectors in your browser, then start the scan and watch live progress.
git clone https://github.com/FlyerFukas/leadradar.git
cd leadradar
py -m pip install -r requirements.txt
py panel.pyYour browser opens at http://127.0.0.1:8765 select a city, optional districts and sectors, then hit "Start scan". Progress streams live and a link to the PDF appears when the run finishes.
Without the panel, from the command line:
py run.py # default weekly rotation
py run.py --limit 5 # quick trial with fewer prospectsRequirements: Python 3.10+ and an internet connection.
LeadRadar works fully without Firecrawl. When enabled, it adds two capabilities:
- Pulls ratings and review counts from directories (yelp.de, gelbeseiten.de, jameda.de…)
- Properly scrapes JavaScript-rendered websites
Important: this repository contains no API keys. The key is read from your own machine at runtime so anyone who clones this project uses their own Firecrawl account and their own credits. Nobody else's credits are ever consumed.
The key is resolved in this order:
# 1) Environment variable
setx FIRECRAWL_API_KEY "fc-your-key" # Windows
export FIRECRAWL_API_KEY="fc-your-key" # macOS / Linux
# 2) Or sign in with the Firecrawl CLI the key is picked up automatically
npm install -g firecrawl-cli && firecrawl loginIf no key is found, LeadRadar prints a notice and continues without Firecrawl.
To disable it entirely, set "firecrawl": { "enabled": false } in config.json.
The free Firecrawl tier is more than enough for this workload (~2 credits per prospect).
| Setting | Description |
|---|---|
city |
Target city (default: Berlin) |
target_leads / discover_pool |
Prospects per report / discovery pool size |
no_website_ratio |
Quota for "no website" prospects (0.6 = 60%) |
category_rotation / area_rotation |
Weekly rotation lists |
category_osm |
Sector → OpenStreetMap tag mapping (add new sectors here) |
chain_blacklist |
Chain / franchise filter |
firecrawl.enabled |
Toggle Firecrawl enrichment |
ai.enabled |
Optional LLM copy polish (OpenAI / Anthropic) |
Weekly rotation or panel selection
↓
Discovery (OpenStreetMap Overpass)
↓
Fingerprint + deduplication (SQLite)
↓
Selection (businesses without a website first)
↓
Website audit (HTTPS · mobile · contact · broken links · errors)
↓
Scoring (HIGH / MEDIUM / LOW evidence only)
↓
Persist + PDF report
A detailed walkthrough of the whole pipeline lives in
docs/LeadRadar_Sistem_Rehberi.pdf (regenerate it with py make_system_guide.py).
The komutlar/ folder contains double-clickable .bat shortcuts (start panel, quick
scan, install the weekly schedule, open outputs). Full command list:
komutlar/KOMUTLAR.md
Weekly automated run (Windows, every Monday at 09:00):
powershell -ExecutionPolicy Bypass -File haftalik_zamanlama.ps1| What | Where |
|---|---|
| Lead report (PDF) | output/LeadRadar_Lead_Raporu_<City>_<date>_<time>.pdf |
| Raw data (JSON) | output/LeadRadar_calistirma_<City>_<date>_<time>.json |
| System guide (PDF) | docs/LeadRadar_Sistem_Rehberi.pdf |
| Memory (SQLite) | data/leads.db |
Every run produces a separate file earlier reports are never overwritten.
output/ and data/ hold real business data and are therefore excluded from the
repository (see .gitignore).
- Business data: OpenStreetMap (ODbL)
- Website audit: the business's own publicly accessible website
- Polite crawling: delays between requests, a single page request per site
- The system never sends messages on your behalf outreach is always your decision
- Only publicly available business information is used
This project started from the logic of the n8n workflow "Local Business Lead Discovery and Enrichment Agent" (Marco's Lead Scout), but was rewritten from scratch as a standalone, deterministic Python application without n8n, OpenAI agents, or an external database. The original workflow's "guardrail" scoring rules were translated into code, which makes the results reproducible and free of hallucination.
MIT