Disha embeds an autonomous conversational AI agent directly onto a live geospatial map. Ask for data, run geodesic spatial analysis, generate zoning scenarios, and produce rich planning reports in natural language.
Traditional GIS requires memorizing nested menus, obscure tool dialogs, and manual geometry conversions. Disha replaces that friction with an agentic engine equipped with real, executable geospatial tools.
Every operation runs real server-side Python math: geodesic ellipsoidal buffers via Shapely + pyproj, DuckDB parquet querying over Overture Maps, OpenStreetMap Overpass extraction, and traffic bottleneck assignment algorithms.
Every tool in Disha maps to verified, tested Python and React source code. Explore the complete capability matrix below.
Multi-turn streaming conversational agent with interrupt/cancel, tool execution reasoning, right-click map context queries, and attachment parsing.
Fly-to bounding boxes, geocoding search, manual polygon/line drawing with undo/redo, OSRM distance measurement, and 7 switchable vector basemaps.
Categorized and graduated data-driven styling with color ramps, dynamic on-map text labeling, interactive collapsible legends, and hover inspection cards.
Shapefile, GeoPackage, KML/KMZ, GPX, and CSV ingestion with auto coordinate detection. Per-layer GeoJSON export, clip-to-view extent, and high-res PNG/PDF map layouts.
Export notes, tables, metrics, and geometry directly into reusable artifacts. Export cleanly to Word, PDF, Markdown, or CSV. The agent re-opens past work seamlessly.
Buffer, convex hull, centroid, area calculation, union, intersection, difference, clip, dissolve, nearest neighbor, and spatial join computed on WGS84 ellipsoid.
OpenStreetMap Overpass, Overture Maps S3 Parquet via DuckDB, Google Earth Engine rasters, WorldPop demographics, India DataMeet, and GTFS feeds.
Zone overlap detection, parcel area breakdowns, and AI scenario generation (TOD, Compact Growth, Green Corridors) scored against real OSM accessibility indices.
Drop in masterplan PDFs or scanned maps: vision-driven automated and manual georeferencing, OCR extraction, and raster feature boundary digitization.
Integrated 360° panorama viewer, coverage street overlay, save-to-artifact inspection shots, and corridor-level road asset review gallery.
Long-form multi-section spatial research reports with live streaming generation, compiling demographic, environmental, and mobility insights.
Save and recall exact map extents, layer visibility states, symbology configurations, and study bounding boxes with one click.
Local folder-based project isolation, automatic last-workspace restoration, non-destructive file tree browser, and zero cloud lock-in.
Environment API key management, local SQLite/disk cache for remote geospatial queries, and strict rate-limiting respecting OpenStreetMap & Nominatim policies.
How Disha handles massive geospatial datasets and transportation math without heavy C++ runtime dependencies.
Queries Overture Places, Buildings, and Transportation layers directly against AWS S3 Parquet releases via embedded DuckDB spatial extensions. Zero local database setup required.
Constructs topologically connected OSM road graphs respecting turn restrictions, runs Dijkstra/A* pathing, and assigns origin-destination matrices to identify network bottleneck links.
By replacing heavyweight GDAL binaries with pure Shapely, pyproj, and Fiona/GeoPandas wheels, Disha packages cleanly into native macOS and Windows desktop installers.
Download official release binaries for macOS and Windows, or clone the repository to run from source.