Sort buffer size per thread in MB. 0 = auto-detect based on RAM.
--tmpdir
path
system temp
Directory for temporary sort chunk files. Use a fast SSD.
--node-store
path
—
Path to mmap file for OSM node coordinates (96+ GB for planet).
--node-cache
bool
true
Cache node store between runs (skip OSM pass 1 on re-runs).
--resume
bool
false
Resume from existing sort chunks in --tmpdir, skipping Phase 1.
--reclaim-source-after-read
bool
false
Delete each Overture Parquet file once its features are in the sort chunks, keeping the disk peak near the release size instead of release + chunks. Only Overture inputs are removed; re-download the release to restore.
Geocoding data output path (requires --output-format geocoding).
--geocoding-format
string
glx
Geocoding format: glx (one index file for geolith search), pg-copy, photon, or pelias.
--pg-index-flavor
tin | gin
tin
Text index in the pg-copy companion DDL. tin needs the tin extension (PlanetScale Postgres, Neki); gin runs on any Postgres.
--wikimedia-importance
path
—
Nominatim's wikimedia-importance.csv.gz. An OSM place with a wikidata tag takes the importance of its Wikipedia article; without the file, each place takes the default of its type. See Importance.
Reads Overture places and addresses and, with --osm-pbf, named OSM settlements, venues, and addresses (all respecting --bbox). The pelias format additionally writes Overture buildings to a <stem>-buildings.ndjson sidecar for a separate buildings index. See the Geocoding Output guide for the .glx file, document shapes and import commands, and the Ecosystem page for how each output reaches a server.