I’m building a scalable delivery platform and need a Python-based method to spot “unknown attack vectors” (UAVs): real residential or business addresses that our current Nominatim geocoder fails to recognise. The task – Query high-risk regions where OpenStreetMap coverage is patchy, surface genuine street-level locations that slip past Nominatim, and return each finding in the JSON shape below: ``` uav: { address: "Somerset East Bosberg Nature Reserve, Blue Crane Route Local Municipality, Eastern Cape, 5850, South Afric", label: "Real house found not in Nominatim", longitude: 25.560463, latitude: -32.703228 } ``` Technical frame • Core language: Python (async preferred for heavy I/O). • Data store: MongoDB for caching queries and results. • External look-ups: you are free to combine secondary geocoders (e.g. Google Maps, local cadastral feeds, postal APIs) and any satellite or crowdsourced dataset that strengthens accuracy; please log the source you ultimately trust for every coordinate pair. Deliverables 1. Clean, well-documented Python code that: – accepts a any country name – flags those missing from Nominatim, – enriches them with the best-guess lat/long, – stores everything in a MongoDB collection. 2. A starter dataset containing at least 50 validated UAV entries spanning several country/region profiles so I can test recall. 3. Read-me explaining setup, environment variables (API keys, rate-limits) and simple reproduction steps. Acceptance I’ll run the script against a fresh batch of addresses; if it consistently returns real-world coordinates (±200 m) for entries Nominatim cannot resolve, the job is done. Feel free to suggest smarter heuristics for defining “high-risk” territories or refining confidence scores—the more resilient the detection, the better. P.S core issue is to reduce search time.