Gregan: Glendora Open-Data Explorer
Gregan is a Glendora, California open-data explorer built on the same Elvis, Robbins, and Groening engine: free public datasets fetched into a DuckDB warehouse, modeled with dbt, and served through a multi-page Streamlit app as maps, charts, and searchable tables. The interesting part is the sourcing. Glendora's data splits three ways, and that shape drives the warehouse: the city's own ArcGIS server for parks, street trees, and zoning; county, state, and federal files filtered down to Glendora for inspections, groundwater, air quality, and wildfire; and a logged DROP list for everything not machine-readable. It is the same reproducible pipeline pointed at a harder scope-a-region-to-one-city problem.
What it proves
Multi-source ingestion narrowed to one small city, pulling three access patterns (city ArcGIS, filtered county and federal files, and Socrata-style APIs) into one tested DuckDB and dbt warehouse.
How it was built
Claude Code on the RECL loop, a reproducible ELT-plus-dbt warehouse baked at build time, with dbt tests on the models.
What is real, and where I was learning
What is real: the ingestion patterns, the dbt models, and the live app all run on Railway. Where I was learning: it deliberately reuses the Elvis architecture, so the work is in the sourcing and the city-scoping, not a from-scratch design.