// the find
cogeotiff/rio-tiler
User friendly Rasterio plugin to read raster datasets.
rio-tiler is a Python wrapper around Rasterio/GDAL that makes it straightforward to read raster data as map tiles, bounding-box crops, or point samples. It's the foundation under titiler and cogeo-mosaic, so if you're building a dynamic tile server or doing geospatial analysis on Cloud Optimized GeoTIFFs, this is the library you'd reach for. Not useful outside the geospatial domain.
Native COG support means HTTP range requests work correctly so you're not downloading entire files to serve a single tile. The reader abstraction is consistent across local files, S3, GCS, STAC catalogs, and Xarray datasets — same tile/part/feature/point API regardless of source. morecantile integration means you're not locked to Web Mercator; EPSG:4326, UTM zones, and custom grids all work. The mosaic_reader handles multi-file compositing with pluggable merge strategies, which would otherwise require significant GDAL wrangling.
The GDAL/Rasterio dependency chain is notoriously painful to install on Windows and in CI without conda or a pre-built wheel — pip install alone frequently fails. XarrayReader is still marked experimental for Zarr and async paths, and the plans/ directory shows the async Zarr reader is not yet merged. Expression evaluation (band math like '(b1-b2)/(b1+b2)') had a security hardening plan opened, which suggests it was executing user-supplied strings unsafely. Stars are low (589) for a library this foundational, which reflects its niche audience but also means community troubleshooting resources are thin.