One .vpk file per region in /vmap (header, sorted index of tiles and
points, then the data), read in place with the index in PSRAM; loose
{dz}/{x}/{y}.vt/.vp files still work. osm_vector.py writes packs
(--pack/--name), reads OSM PBF extracts for a box (--pbf/--bbox, via
pyosmium) and picks label languages (--lang). Map tools > Vector regions
lists the packs on the card: show one on the map, delete it. No
on-device download: packs are made with the tool.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
osm_vector.py --dem adds contours from the Terrain Tiles on AWS Open Data
(SRTM, ~25 m; tools/maps/dem.py, standard library only: PNG decode,
smoothing, marching squares, line joining). Every 100 m from zoom 12,
every 20 m from zoom 15; the credit names the elevation source.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Hiking routes are stored per stretch with the waymark colours of every
route along it and drawn as side-by-side stripes on a white band (half
the route ways in the Tatra sample carry two or more routes), less
simplified; demanding / alpine paths are dotted. Named points (places,
peaks with height, huts, passes, springs, viewpoints, caves) come in
small .vp files and are placed by priority as a layer over the tiles,
clear of each other and of the map controls.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Points as zigzag varint deltas, a byte length per feature to skip it,
natural land cover simplified harder: the Tatra sample goes from 4.4 MB
to 2.3 MB, panning at z14 from ~63 to ~45 ms per tile on the L2.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
tools/maps/osm_vector.py turns Overpass JSON into VT2 tiles (data zooms
10/12/14, per-feature bbox, simplified); VectorTileProvider draws them
into 256 px tiles (scanline polygons, thick lines with round joins) with
a small PSRAM cache of data files. Behind Map tools > Vector map (test),
raster where there is no vector data.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
JakubandClaude Opus 5.5