Convert addresses to map coordinates without any internet connection. Powered by OpenAddresses open data and DuckDB Spatial, the Offline Geocoding module enables single-address lookup, bulk batch processing, and reverse geocoding — all on-device in denied environments.
In DDIL (denied, degraded, intermittent, limited) environments, address-to-coordinate conversion is typically impossible without internet access. Tactical AI DuckDB eliminates this dependency by embedding a local OpenAddresses dataset pre-indexed in DuckDB Spatial for sub-second lookups against hundreds of millions of address records — entirely offline.
An emergency management team receives a list of 500 addresses in an affected area and needs to plot them on the map for resource allocation. Using Batch Geocoding, all 500 addresses are geocoded in seconds against the local OpenAddresses dataset, creating map markers with occupancy and address data — with no connectivity required.
The geocoding engine is powered by OpenAddresses — the world's largest open dataset of address points collected from authoritative government sources. Regional datasets are downloadable in advance for pre-staging before deployment.
Three geocoding modes cover all field use cases from single address lookups to large-scale batch processing:
| Mode | Description | Best For |
|---|---|---|
| Interactive | Type address, get coordinate instantly with map preview | Single lookups, navigation waypoints |
| Batch (CSV) | Upload CSV with address column, download coordinates | Pre-staging large address lists |
| Reverse | Tap map or enter coordinates, get nearest address | Identifying locations, incident reporting |
All geocoding results are available in multiple coordinate reference systems to match your operational workflow:
Large-scale batch geocoding leverages DuckDB's columnar processing engine. Input CSV files with thousands of addresses are processed using parallel SQL joins against the OpenAddresses spatial index, returning results in seconds rather than minutes.
| Dataset Size | Estimated Processing Time | Device |
|---|---|---|
| 100 addresses | < 1 second | Any Android 8+ device |
| 1,000 addresses | 2-5 seconds | Any Android 8+ device |
| 10,000 addresses | 15-30 seconds | Snapdragon 8-series recommended |
| 100,000 addresses | 2-5 minutes | High-performance device |
When connectivity is available, the geocoder can optionally augment OpenAddresses results with online services for higher accuracy or coverage in regions where local data is limited:
The AI Agent integrates with the geocoding engine, enabling natural language location resolution. When you ask "Navigate to 123 Main Street," or "Mark the building at the intersection of Oak and Elm," the AI automatically resolves the address to coordinates using the offline geocoder and performs the requested action.
During route planning, the AI Agent geocodes multiple waypoints from a verbal briefing: "Plan a route from the staging area at 100 Industrial Drive to the checkpoint at 450 Highway 35, then to the forward operating base at grid 37SED 123 456." All three location types (address, address, MGRS) are resolved automatically.
Geocoding datasets are downloaded per region before deployment and stored on-device. The Dataset Manager shows download status, storage usage, and allows updating when connectivity is available.
| Region | Approx. Download Size | Record Count |
|---|---|---|
| Single US State (e.g. Texas) | ~200 MB | ~12M addresses |
| Full US Coverage | ~4 GB | ~150M addresses |
| European Country | 100-500 MB | 5-40M addresses |
| Custom AOI | Variable | Area-dependent |