Case study 04 / Windows local web app

PinHarbor

A Docker-free Go and Playwright workflow with clearer presets, resilient supervision, local job history, and truthful telemetry.

Conceptual artwork / product evidence follows

Product direction, interface, workflow integration, telemetry, and verification by Kirsten Trimaley; built on the attributed upstream foundation.

I use AI to explore and accelerate, but I own the product decisions, the verification, and the responsibility for what ships.

The real product

Built to be used,
not just viewed.

A Windows-local Google Maps business collector with completion-driven runs, honest live progress, and focused CSV exports.

Complete application frame · source aspect ratio preserved · privacy-safe data

Built a compact workflow around Auto complete, Quick, Balanced, Deep, and Custom collection modes.
Connected the interface to real job CSV rows and thread-safe progress, then added a supervised Windows launcher and recovery checks.

Progress cannot promise a fixed total because new places may be discovered while the run is active.
Existing jobs, SQLite state, CSV exports, and upstream MIT attribution must remain intact.

Keep active jobs below 100 percent and label ETA honestly as calculated, finalising, stopped, or bounded by configured limits.
Run directly with Go and Playwright on Windows instead of requiring Docker or WSL.

Go tests, focused runner and web tests, browser checks, export-field coverage, and supervisor recovery passed.
After a forced process stop, the launcher replaced the listener and the health endpoint returned ok.

Foundation and attribution: PinHarbor retains the MIT license and attribution for the gosom/google-maps-scraper project it builds on.

Keep the collector focused on responsible, operator-directed research and maintain upstream compatibility as dependencies change.