Processing

Georeference rasters with affine, polynomial or TPS transforms, RMS residuals and direct COG export via GDAL/WASM.

Drive an AI agent that edits layers, runs SQL/Python/MapLibre, executes processing tools and searches STAC; code approval required.

Run most geoprocessing tools in-browser via WASM with an area-of-interest picker, source filter and shareable ?tool= links.

Run spatial SQL across DuckDB-WASM, PostGIS (PGlite) and Sedona engines, with live layers and CSV/GeoParquet export.

Script the map from a Pyodide console with on-demand packages, run-selection editing and .py open/save.

Documents an analysis step by step with text, code, results, and visualizations for reproducible, shareable workflows.

Build dashboards with eight widget types, KPI aggregations, editable grids and DuckDB query-layer sources.

Re-run or edit-and-rerun past executions and export a run as JSON or Python, with inputs, status, duration and outputs.

Adds authenticated Earth Engine images, image collections, features or geometries as normal XYZ layers from the Python API. Authentication and Cloud project quota stay in the kernel; map IDs and tile URLs expire, project files keep only layer references, and live widgets retain runtime tile tokens only for rendering. Map.to_project() and Map.save_project() always redact Earth Engine tile URL tokens, including with keep_credentials=True; that option preserves unrelated credential fields only, HTML exports are redacted too, and credentials are never persisted.

Geocode a CSV address column into a point layer that keeps your original fields plus coordinates and match score.

Run the same vector tool across many selected layers in one pass instead of repeating it layer by layer.

Build and save directed processing graphs on a visual canvas with explicit input/output nodes, labelled ports, typed inspectors, bounded import/export, and Run/Cancel controls. Batch Tools remains a separate Meratyn Toolbox action, and Whitebox/WASM tools are listed only when an executable catalog adapter is available.

Re-run or fine-tune any past execution, and review the generated JSON or Python script before copying it.

Choose per run whether vector tools execute with Turf.js in the browser, GeoPandas on the sidecar, or Pyodide.

Transfer attributes between layers by spatial relationship instead of matching table keys.

Split off a random subset of features for sampling, QA or training data.

Generate grids of evenly sized cells over a drawn extent or around a layer's features.

Draw antenna sector wedges from point attributes such as azimuth, radius and beamwidth.

Fill a layer geometry, layer extent, map viewport, or manual bounding box with H3, S2, A5, DGGRID, or DGGAL cells at a selected or automatic resolution; DGGAL grid and DGGRID cell type, projection, and aperture options are available. H3 uses DuckDB-WASM, S2/A5 run in the browser with bounded A5 center-sampling, and DGGRID/DGGAL use bundled WebAssembly engines.

Aggregate point features into H3, S2, A5, DGGRID, or DGGAL cells with count, sum, mean, minimum, or maximum; a numeric field is required except for count.

Compact complete sibling sets or expand hierarchical H3, S2, A5, or DGGAL cell layers to a target resolution.

Compute per-point speeds along GPS tracks from timestamps, grouped by moving object.

Flag where tracked objects dwell, using a distance radius and a minimum stay duration.

Find pairs of tracked objects that meet within both a distance and a time window.

Locate statistically significant hot and cold spots of a numeric attribute.

Classify grid cells over time as intensifying, diminishing or persistent hot spots using a space-time cube.

Estimate a continuous density surface from weighted points as a gridded heatmap layer.

Normalize, weight, and combine numeric fields into an auditable suitability index.

Route through waypoints in field order via live Valhalla routing, snapping off-road points to roads.

Run YOLO ONNX models over GeoTIFFs or geotagged photos in the browser; detections land as georeferenced layers.

Segment every object in a GeoTIFF automatically with SlimSAM and get one georeferenced polygon layer.

Typed SamGeo/SAM3 sidecar workflow for Processing → Meratyn Toolbox → AI Segmentation text and automatic prompts; reprojects GeoJSON before publishing a layer. The separate Plugins → SamGeo panel adds SAM3 point/box prompts and SAM2 automatic masks, with model/backend controls. A reachable service running the operator-managed [samgeo-runtime] extra on Python 3.12+ with segment-geospatial[api,samgeo3]==1.4.2 is required; operators supply model weights, accept their terms, and set both readiness flags.

Re-encode plain GeoTIFFs as tiled Cloud Optimized GeoTIFFs with overviews, fully in the browser.

Remap raster cell values with a classification table to turn continuous surfaces into classes.

Open any processing tool from a URL with its parameters prefilled, ready to share with teammates.

Search current disaster reporting through the managed Tavily news route with a fixed single-request cost, bounded query and result sizes, and safe failures that never expose operator credentials or retain provider URLs in serializable errors.

Convert rasters into tiled Cloud Optimized GeoTIFFs with overviews and seven compression codecs. Browser handles GeoTIFF input.

Convert vector datasets to Hilbert-sorted FlatGeobuf with a packed spatial index for fast cloud-native access.

Convert vector datasets to GeoPackage for lossless exchange with QGIS, ArcGIS, and other GIS tools.

Convert vector datasets to zipped ESRI Shapefiles, truncating field names to ten characters as the format requires.