ogee: Export OSM Road Networks as Graphs from Your Browser

When working with road networks — for shortest-path algorithms, traffic simulations, or operations research projects — the first step is always the same: get graph data. Tools like OSMnx do this well from Python, but sometimes you want to grab a quick subgraph of the area you are looking at right now without leaving the browser. That is what ogee (OSM Graph Export Extension) does.

What it does

ogee is a Chrome extension that adds an "Export" button to openstreetmap.org. Click it, and a sidebar panel appears with bounding-box coordinates pre-filled from your current map view. Hit "Generate Graph", and the extension fetches road data from the Overpass API, converts it into a directed graph, shows a canvas preview, and lets you download the result in one of four formats: JSON, GraphML, CSV, or LaTeX TikZ.

The graph is directed: one-way streets produce a single edge, two-way streets produce edges in both directions. Edge weights are real-world distances in kilometers, computed via the Haversine formula. Each edge also carries the OSM highway type (e.g. residential, primary) and the street name where available.

Architecture in 30 seconds

The extension follows Chrome's Manifest V3 message-passing pattern:

Content script — injected into openstreetmap.org, manages the UI (nav button, sidebar panel, coordinate inputs, canvas preview).
Background service worker — receives bounding-box coordinates, queries the Overpass API, converts raw OSM elements into a node-link graph, and serializes to the requested export format.
Popup — alternative lightweight UI accessible from the extension icon.

For larger areas the background worker automatically tiles the bounding box into ~5 km chunks, fetches them sequentially (to stay within Overpass rate limits), and merges the results with deduplication.

All pure logic — graph conversion, format serialization, bounds validation, tiling — lives in a shared lib/graph-utils.js module, which is covered by a Jest test suite.

Using the output

NetworkX (Python)

The JSON export uses NetworkX's node-link format, so importing is a one-liner:

import json, networkx as nx
from networkx.readwrite import json_graph

with open("osm-graph.json") as f:
    G = json_graph.node_link_graph(json.load(f))

print(f"{G.number_of_nodes()} nodes, {G.number_of_edges()} edges")

GraphML works just as directly:

G = nx.read_graphml("osm-graph.graphml", node_type=int)

Plotting

Since every node carries lat and lon attributes, plotting the network on a 2D plane is straightforward:

import matplotlib.pyplot as plt

pos = {n: (d["lon"], d["lat"]) for n, d in G.nodes(data=True)}
nx.draw(G, pos, node_size=2, width=0.3, arrows=False)
plt.axis("equal")
plt.show()

LaTeX TikZ

The TikZ export produces a standalone .tex file that compiles directly with pdflatex. Node positions are projected from geographic coordinates to a local coordinate system scaled to 10 cm on the longer axis. Bidirectional edges are drawn once to keep the output clean.

\documentclass[tikz]{standalone}
\usepackage{tikz}
\begin{document}
\begin{tikzpicture}[
  vertex/.style={circle, fill=black, inner sep=0pt, minimum size=1.5pt},
  edge/.style={draw, thin, black!40},
]
  \node[vertex] (n0) at (0.0000,5.0000) {};
  \node[vertex] (n1) at (5.0000,10.0000) {};
  ...
  \draw[edge] (n0) -- (n1);
  ...
\end{tikzpicture}
\end{document}

This is useful when you need a road network figure in a paper or thesis and want full control over styling. Since it is just TikZ, you can adjust colors, line widths, add labels, or overlay additional annotations directly in LaTeX.

Limitations

The Overpass API has rate limits and can be slow for very large areas. ogee mitigates this with tiling, but city-scale exports can still take a while.
Minor road types (footways, cycleways, service roads, tracks) are excluded by default — the focus is on the drivable road network.
The extension only runs on openstreetmap.org. It does not work on other map providers.

For heavy-duty analysis involving entire cities or custom filtering, OSMnx remains the better tool. ogee fills a different niche: quick, visual, browser-native graph extraction for the area you happen to be looking at.