Tile Rendering

Map tile rendering is the second compute type on Hive. The tile service is MapLibre GL JS and flutter_map compatible — swap one URL and any mapping application runs on Hive tiles.

Tile API

GET /v1/tiles/{h3_cell}/{zoom}/{style}.{format}

Examples

# OSM bright style, Rome, zoom 15
curl -o tile.png https://gns-browser-production.up.railway.app/v1/tiles/871e9a0ecffffff/15/osm-bright.png

# Dark mode
curl -o dark.png https://gns-browser-production.up.railway.app/v1/tiles/871e9a0ecffffff/15/dark.png

# Satellite imagery
curl -o sat.png https://gns-browser-production.up.railway.app/v1/tiles/871e9a0ecffffff/14/satellite.png

Available styles

Style	Source	API Key Required
osm-bright	OpenStreetMap	No
osm-standard	OpenStreetMap	No
dark	CartoDB Dark Matter	No
terrain	Stadia/Stamen	No
satellite	ESRI World Imagery	No

Response headers

Every tile response includes Hive provenance headers:

Header	Description
X-Hive-Worker	Worker public key or tile-proxy
X-Hive-Cell	H3 cell of the computation
X-Hive-Epoch	GNS epoch number
X-Hive-Proof	SHA-256 hash of the tile data
X-Hive-Cache	HIT or MISS
X-Hive-Cost	Cost in GNS (0.00001 hit, 0.0001 miss)

Cache strategy

MobyDB is the cache. The composite key (H3 Cell, Epoch, Style, Zoom) IS the cache key.

• Cache hit: serve from memory, <1ms, 0.00001 GNS
• Cache miss: proxy from upstream, cache locally, serve to requester, 0.0001 GNS
• Epoch invalidation: tiles refresh when the epoch advances and source data has changed

MapLibre GL JS integration

const map = new maplibregl.Map({
  container: 'map',
  style: {
    version: 8,
    sources: {
      'hive-tiles': {
        type: 'raster',
        tiles: ['https://gns-browser-production.up.railway.app/v1/tiles/871e9a0ecffffff/{z}/osm-bright.png'],
        tileSize: 256,
      }
    },
    layers: [{
      id: 'hive-layer',
      type: 'raster',
      source: 'hive-tiles',
    }]
  }
});

flutter_map integration

TileLayer(
  urlTemplate: 'https://gns-browser-production.up.railway.app/v1/tiles/871e9a0ecffffff/{z}/osm-bright.png',
)

Compound jobs with tiles

Tiles can be part of a compound compute job. The inference step finds locations; the tile step renders the map:

{
  "steps": [
    {"id": "think", "type": "inference", "messages": [...]},
    {"id": "show",  "type": "tile_render", "depends_on": "think",
     "center_cell": "871e9a0ecffffff", "zoom": 15, "style": "osm-bright"}
  ]
}

The tile step reads locations from the inference output and generates annotation metadata.

Competitive pricing

Provider	Cost per 1,000 tiles	Provenance	Data Sovereignty
Mapbox	$0.60–$5.00	None	No
Google Maps	$7.00	None	No
Hive Maps	$0.10	Merkle proof per tile	Yes (H3 cell binding)

Tile grid endpoint

GET /v1/tiles/grid/{cell}/{zoom}/{style}?rings=1

Returns a JSON array of tile URLs covering the cell and surrounding rings.

Tile stats

GET /v1/tiles/stats?hours=24

Returns hourly aggregated tile serving statistics.