Offline maps & routing in Docker
Corviont runs MapLibre UI, vector tiles, Valhalla routing, and SQLite geocoding entirely offline on edge or on-prem devices - with a local map updater shipping next.
Corviont runs MapLibre UI, vector tiles, Valhalla routing, and SQLite geocoding entirely offline on edge or on-prem devices - with a local map updater shipping next.
A drop-in mapping stack for edge/on-prem deployments - predictable footprint, local APIs, and a MapLibre UI. Each card below is one piece of the system.
Packaged as a single PMTiles file and served via a local tile endpoint - no external tile servers.
Valhalla container exposing an HTTP API for offline routing between arbitrary points.
SQLite database built from Nominatim data, powering forward and reverse geocoding via a lightweight API.
MapLibre frontend wired to these local endpoints.
Corviont is built for edge/on-prem deployments - fleets, gateways, kiosks, vessels, and field devices - where WAN is slow/unreliable, air-gapped, or egress costs matter.
We benchmarked it on real hardware and grouped regions into coverage tiers (RAM, download size, on-disk). Each tier lists a few example regions to help you pick the closest match.
CityVienna, Berlin, DC
|
StateAustria, Bavaria, Colorado, Texas
|
LargeSpain, Poland, UK, California
|
ComplexGermany, Western US
|
|
|
RAM usage
p95 of Corviont containers, OS overhead not included
|
200 MB
|
500 MB
|
1.5 GB
|
3 GB
|
|
Download size
Tiles + routing + search, does not include container images (0.5 GB)
|
200 MB
|
2 GB
|
4 GB
|
8 GB
|
|
On-disk data
Tiles + routing + search, does not include container images (1.5 GB)
|
300 MB
|
3 GB
|
6 GB
|
12 GB
|
CityVienna, Berlin, DC
|
|
RAM usage
p95 of Corviont containers, OS overhead not included
200 MB
|
|
Download size
Tiles + routing + search, does not include container images (0.5 GB)
200 MB
|
|
On-disk data
Tiles + routing + search, does not include container images (1.5 GB)
300 MB
|
StateAustria, Bavaria, Colorado, Texas
|
|
RAM usage
p95 of Corviont containers, OS overhead not included
500 MB
|
|
Download size
Tiles + routing + search, does not include container images (0.5 GB)
2 GB
|
|
On-disk data
Tiles + routing + search, does not include container images (1.5 GB)
3 GB
|
LargeSpain, Poland, UK, California
|
|
RAM usage
p95 of Corviont containers, OS overhead not included
1.5 GB
|
|
Download size
Tiles + routing + search, does not include container images (0.5 GB)
4 GB
|
|
On-disk data
Tiles + routing + search, does not include container images (1.5 GB)
6 GB
|
ComplexGermany, Western US
|
|
RAM usage
p95 of Corviont containers, OS overhead not included
3 GB
|
|
Download size
Tiles + routing + search, does not include container images (0.5 GB)
8 GB
|
|
On-disk data
Tiles + routing + search, does not include container images (1.5 GB)
12 GB
|
Here’s what we’re building next to support larger coverage and real-world deployments:
🧩 Local map updater
A small background service that pulls new map bundles, verifies them, and switches the active dataset without downtime.
🗂 Custom overlays
Load your own POIs, geofences, or operational layers (GeoJSON) on top of the base map, rendered directly in the UI.
📍 Richer geocoding output
Better address results with house numbers, and optional geometry for streets and areas (not just centrepoints) in forward and reverse search.
🧱 More edge platforms & targets
First-class integrations for Portainer and Mender, plus deployment examples for K3s/Kubernetes and edge runtimes on AWS and Azure.
Have a real deployment in mind? Share your region and we’ll follow up to scope a paid pilot.
The form has been successfully submitted.
