GeoPrints

Overview

Built for: Android users who want an automatic personal travel and location journal without sending their location history to a cloud service.

Main features

• Automatic background tracking

  An Android foreground service samples location with separate profiles for walking, cycling, driving and staying still. Writes are buffered, and an open stay resumes correctly after a restart.

• GPS noise filtering

  Mock locations, inaccurate fixes and physically impossible speeds are dropped. A sudden far jump is held until the next fix confirms whether it was real movement or a spike.

• Stay detection

  A stay begins when recent fixes over several minutes fit within an accuracy-aware radius, and ends only after consecutive fixes fall outside the exit radius, so brief GPS drift does not split a visit.

• Today's Journal

  A chronological timeline of stays and movements for the day, with durations, distance for each movement and a total count of stays.

• Gallery photo matching

  Photos taken today are attached to the timeline event closest in time. The app reads only timestamps and file paths, and the photos never leave the device.

• Map Explorer with heatmap

  Cleaned routes split by activity, clustered stay and photo markers, follow mode, dark and light basemaps and a heatmap of frequently visited areas. Viewed map tiles are cached on the device.

• Place labels

  Frequently visited places can be labelled Home, Office, Gym, Food or a custom name, and those labels then replace raw addresses across the timeline.

• History calendar and search

  A month calendar marks recorded days and opens each day's activity. Search covers past locations with quick filters such as stays longer than 15 minutes, food and coffee.

• Life balance and daily story

  A rolling 7-day split of time at home, at the office, in transit and elsewhere, plus a rule-based written summary of each day generated on the device.

• Data ownership controls

  Export the full history as JSON and GPX through the share sheet, auto-delete records older than 30 days, clear the map cache or delete all data.

Development and architecture

Privacy by architecture

I chose to have no backend, no accounts and no analytics at all. Location data stays in the on-device database unless the user starts an export, so privacy does not depend on a policy page alone.

Tested tracking core

The tracking and route-cleaning code is covered by 20 unit tests, including cases for mock fixes, accuracy limits, speed limits, spikes and stay entry and exit.

Battery-aware sampling

Sampling intervals change with the motion mode reported by Android activity recognition, with a short fast-sampling window to catch the moment a user leaves a stay.

On-device route cleaning

An earlier build snapped routes using a public OSRM server. I replaced it with on-device smoothing that breaks lines at time and distance gaps, splits them by activity and removes redundant points, so recorded routes are no longer sent to an outside server.

Shared database across isolates

One Isar schema list is used by both the UI and the background service isolate, with stays, location points, place labels and a geocode cache keyed at roughly 50 meters to avoid repeated lookups.

Play policy compliance

A prominent location disclosure and a permissions hub explain each permission before it is requested. No native permission prompt fires on launch, and restart on boot only happens when background location is granted.

Large exports

JSON and GPX 1.1 exports are written in pages of 5,000 records, so long histories can be exported without loading everything into memory at once.

Technology

• Flutter
• Dart
• Android
• Riverpod
• Isar
• flutter_map
• OpenStreetMap
• Android Activity Recognition
• GitHub Actions
• Tailwind CSS
• JavaScript