A demo with a synthetic dataset
Find every garage sale on sale day.
On a community sale day, the list of stops is often one long blog post. This demo shows another way: an AI pipeline that turns a sale list into structured records, and a phone-friendly map you can filter by category or keyword between stops.
The 30 stops here are made up. They sit on real West Seattle blocks so the map reads right, but no stop is a real sale or a real household.
How this works
Step 1
Read the list
A sale day usually lives in one source: a blog post, a PDF, or a community calendar. The ingest script strips it to plain text.
Step 2
Structure it
Claude turns each entry into a record with address, hours, categories and items, and a schema check rejects bad rows. So far this has only run on a three-row sample in mock mode.
Step 3
Filter and favorite
Filter by category or keyword, star the stops you want, and share the link with whoever is riding along.
See a sample input, its JSON output and the validation rules