Timeline Builder

Google Maps lets someone share their live location with you — but that’s all you get. A dot on a map, right now. No history, no timeline, no way to look back. If you want a timeline of a shared location, you’re on your own.

This project builds that missing timeline by scraping the shared location every two minutes and feeding it into Dawarich — an open-source Google Timeline alternative running on your own hardware.

How It Works

A headless Chromium browser runs inside a Docker container, authenticated to a Google account with shared location access. Every two minutes, it hits the Google Maps location sharing endpoint, extracts coordinates, and pushes them to Dawarich via the OwnTracks API.

Dawarich handles storage, geocoding, and visualisation — maps, heatmaps, stats, and export.

Architecture

Google Maps Shared Location — polled every 120 seconds
Scraper Container — Python + Playwright + headless Chromium
Dawarich — Rails web UI + API, with Redis for job queuing
Sidekiq Workers — background processing
PostgreSQL + PostGIS — all location history, backed up daily

Features

Automated collection — every 2 minutes, zero manual input
Self-hosted — all data stays on your hardware
Full Dawarich UI — maps, heatmaps, stats, export
Email alerts — notification if the scraper loses its session
Zero-touch — runs unattended, survives host reboots without re-auth
Automated backups — daily database dumps to Google Drive
HTTPS-only — Nginx TLS sidecar, no exposed HTTP ports

Tech Stack

Scraper: Python 3.12, Playwright, headless Chromium
Application: Dawarich (Ruby on Rails, Sidekiq)
Database: PostgreSQL 17 with PostGIS
Cache/Queue: Redis
Infrastructure: Docker Compose, ZeroTier

Challenges

Google Auth in a headless browser. Google aggressively blocks automated logins — it checks browser fingerprints, viewport dimensions, user agent strings. The scraper runs a full Chromium instance, not a lightweight HTTP client, and mimics a real browser session. Auth is done manually once, then the browser profile persists cookies across container restarts.

Reverse-engineering the response format. Google’s location sharing isn’t a documented API. The scraper intercepts the data that Google Maps loads when you view a shared location. Figuring out which nested array contains the coordinates, timestamp, and accuracy was trial and error with a lot of JSON staring.

Keeping the session alive. The browser profile lives on a Docker volume. If it gets corrupted or Google invalidates the session, the pipeline stops. The scraper monitors its own health and sends an email alert if collection stops. Sessions have survived weeks including host reboots.

Chromium leaks memory. Headless Chromium running Playwright accumulates memory over time — hit 2.4 GB and OOM’d after five days of continuous operation. The pragmatic fix: a 2 GB Docker memory cap as a ceiling, plus a daily scheduled restart at 3 AM. One 30-second gap per day, memory stays bounded. Root cause investigation is backlogged — likely the browser context not fully releasing resources between scrape cycles.

HTTPS reverse proxy. Dawarich is a third-party Rails app — can’t bake TLS into it. An Nginx sidecar terminates HTTPS, but Rails generates redirect URLs using its internal protocol. After login, it would redirect to http:// without the port, breaking the flow. The fix was a combination of proxy_redirect rules, X-Forwarded-Proto headers, Origin header clearing for CSRF compatibility, and passing the port in the Host header. Fiddly, but now it’s a reusable pattern.

Limitations