Command line

The roadstyle command turns a road file into a map or a page, without writing Python.

roadstyle INPUT [options]          # INPUT: GPKG, GeoJSON, Shapefile, … with a road-class column
roadstyle studio [streamlit args]  # the Studio workbench

Each styling flag mirrors a Python keyword (Every parameter); a flag you leave out keeps the keyword's default and your settings.

Output

flag	default	what
-o, --output PATH	input name + the format's extension	output file
-f, --format FMT	web	see Formats
--version		print the version
-h, --help		print every flag

Formats

-f	writes	extension	Python equivalent
web	a self-contained MapLibre map	.html	render_edges(g).save(...)

Only the styling, filtering and data-colour flags apply to every format; the web flags below need -f web.

Flags

flag	Python keyword	what
Input
--highway-col COL	highway_col	road-class column (default highway)
--layer NAME	load_edges(layer=)	layer of a multi-layer file, e.g. a GPKG
Styling
--palette {amber,carto,highsat,mono}	palette	class palette (default amber)
--basemap KEY	basemap	base map (default from settings: ofm_positron)
--tooltip COL [COL …]	tooltip	columns in the hover tooltip
Filtering
--include TYPE [TYPE …]	include	keep only these classes
--exclude TYPE [TYPE …]	exclude	drop these classes
Colour by data
--color-by COL	color_by	colour by this column
--colors JSON	colors	categorical map, e.g. '{"a":"#f00"}'
--cmap NAME	cmap	numeric ramp, e.g. viridis
--vmin N, --vmax N	vmin, vmax	ramp range
--width-by MIN MAX	width_by	width in px follows the numeric --color-by
Web map (-f web)
--page {dashboard,street-view}	render_dashboard / render_street_view	a ready-made page instead of the plain map
--dashboard JSON	dashboard=	with --page dashboard: the dashboard's options, for example the parts of its panel (Dashboard options)
--view-3d	view_3d=True	tilted camera and 3D bridge decks
--pitch DEG, --bearing DEG	pitch, bearing	starting camera
--no-arrows	arrows=False	no one-way arrows
--no-labels	labels=False	no street names
--no-filter	filter_control=False	no effect: the floating Roads box is gone (the dashboard's roads tool is the road filter)
--no-basemap-switcher	basemap_switcher=False	no base-map button
--tiles / --no-tiles	tiles=True / tiles=False	--tiles is the default: the roads are embedded MLT vector tiles, as in render_edges. --no-tiles writes inline GeoJSON roads
--no-compress	compress=False	plain JSON data instead of gzip

--page

page	builds
dashboard	the map, the standard toolbox and an empty panel. --dashboard gives the panel its parts (Dashboards)
street-view	the map + Google Street View of the clicked road

--page needs -f web. For street-view only:

flag	keyword	what
--layout {beside,below}	layout	Street View right of the map (default) or under it
--panel-width PCT	panel_width	Street View's share of the window, 20-80 (default 42)
--no-resize	resizable=False	no draggable divider

Street View needs the page served over http(s): python -m http.server, then open http://localhost:8000/map.html. See Google Street View.

roadstyle-levels

The level step with files you can read and edit (Which road is on top):

roadstyle-levels make SOURCE FOLDER [--query SQL] [--geometry COL] [--id-col edge_id] [--band-col COL] [--order priority|class|COL]
                    [--band-dist 10] [--near-rules] [--engine python|rust]     # roads.parquet, pairs.csv (edits.csv once)
roadstyle-levels solve FOLDER [--auto-ends] [--max-positions N] [--engine python|rust]    # levels.csv
roadstyle-levels edit FOLDER [--port 8780]                                                # the level editor

The Rust program roadstyle

The Rust program has no Python. Its options have the names of the render_edges keywords: --color-by lanes is color_by="lanes", --no-arrows is arrows=False. A value that is JSON is read as JSON. All commands are in The roadstyle program.

command	does
roadstyle page ROADS OUT.html [--palette P] [--profile P] [--settings FILE] [--api-key KEY] [--street-view-key KEY]	the page in one file, as render_edges
roadstyle sources ROADS OUTDIR [--format geojson\|geoparquet] [--keep COL,COL] [--name NAME]	stage 1: the style and one file for each source
roadstyle tiles STAGE1DIR OUT [--format pmtiles\|xyz] [--tile-format mlt\|mvt] [--keep COL,COL] [--base-url URL]	stage 2: the vector tiles
roadstyle levels make EDGES FOLDER [--id-col C] [--band-col C] [--order O] [--band-dist 10] [--head-m 5] [--near-rules] [--highway-col C] [--layer-col C] [--bridge-col C] [--tunnel-col C] [--junction-col C] [--boundary AREA.geojson]	a level area's input. --boundary keeps the edge of the area, which the editor draws as a dashed line
roadstyle levels solve FOLDER / levels edit FOLDER [--port 8780] [--page DIR]	solve the area; the level editor (roadstyle serve FOLDER is the same without the check of the area)
roadstyle design edit ROADS [--design FILE] [--port 8781] [--page DIR]	the design editor
roadstyle snapshot PAGE.html OUT.png [--center LON,LAT] [--zoom Z] [--pitch P] [--bearing B] [--width 1200] [--height 800] [--scale 1] [--settle 2.5] [--timeout 40]	a PNG of a saved page

roadstyle-mcp

Starts the MCP server: three tools for AI agents. Needs pip install "roadstyle[mcp]".

roadstyle studio

Opens the Studio workbench in the browser. Everything after studio goes to streamlit run, e.g. roadstyle studio --server.port 8502. Needs pip install "roadstyle[studio]".

Examples

roadstyle edges.gpkg -o map.html --basemap ofm_dark
roadstyle edges.gpkg --include motorway trunk primary -o major.html
roadstyle edges.gpkg --color-by maxspeed_kmh --cmap viridis --width-by 1 6
roadstyle edges.gpkg --page street-view --layout below --panel-width 50
roadstyle edges.gpkg --page dashboard -o dashboard.html
roadstyle edges.gpkg --page dashboard -o dashboard.html \
  --dashboard '{"panel": {"sections": [{"id": "q", "part": "query", "label": "Query"}, {"id": "road", "part": "road_card", "label": "Selected road"}]}}'

The Rust program takes the same JSON with the keyword --dashboard, and the template with --template dashboard:

roadstyle page edges.gpkg dashboard.html --template dashboard --dashboard '{"panel": {"sections": [{"id": "q", "part": "query"}]}}'

See also: Every parameter · Get started