NYC apartment search: share-site pipeline

Last updated: Oct 21, 2026, 1:22:18 AM

Table of Contents

Pipeline Charts 📈

• Chart List

• Chart: Monthly cost vs. commute

• Chart: Monthly cost breakdown

• Chart: Buildings by neighborhood

• Chart: Building map

Pipeline Dataframes 📊

• Dataframe: gonzalorrd/nyc-apartment-search:buildings - Buildings
• Dataframe: gonzalorrd/nyc-apartment-search:neighborhoods - Neighborhoods
• Dataframe: gonzalorrd/nyc-apartment-search:units - Units

Pipeline Specs

Pipeline Name	NYC apartment search: share-site pipeline
Pipeline ID	gonzalorrd/nyc-apartment-search
Maintainer	Gonzalorrd
Contributors
Repository	Gonzalorrd/nyc-apartment-hunter
Pipeline Web Page	Pipeline Web Page
Date of Last Code Update	2026-10-21 01:22:04
OS Compatibility	Windows, Linux
Linked Dataframes	gonzalorrd/nyc-apartment-search:buildings gonzalorrd/nyc-apartment-search:units gonzalorrd/nyc-apartment-search:neighborhoods

Build Commands:

uv run doit

This is the documentation for the pipeline behind the apartment search page. It started as a data-pipeline class exercise: automate the flow from a private local database to a public, visual page, using GitHub Actions.

laptop:  apartments.db --apartment-hunter publish--> site_data/snapshot.json --git push-->
Actions: validate -> enrich -> transform -> render -> chartbook -> GitHub Pages

Step	What it does	Code
publish	Writes an allowlisted snapshot: buildings, units, commutes and risks only. Notes are redacted; contacts, tours, tasks and evaluations never leave the laptop.	publishing.py
validate	Data contract (the app’s Pydantic schemas, foreign keys, unique ids) and privacy gate (allowlist, redaction, email/phone patterns). Fails the build on any problem.	share_pipeline.validate
enrich	Map coordinates from the US Census Geocoder, cached between runs. Matches on a different street are rejected. Outages never block a deploy.	share_pipeline.enrich
transform	Ranks units with the app’s own scoring.rank_unit and writes parquet tables. Freshness depends on the date, so a daily scheduled run keeps it accurate.	share_pipeline.transform
render	Plotly charts, the phone-first landing page and CSV downloads.	share_render.py
chartbook	This documentation site, with a page for every chart and table.	share/chartbook.toml

Run everything locally with uv run doit from the repository root, and list the tasks with uv run doit list. The output is written to share/public/.

Limits: “checked” means a public source was looked at on that date. It is not a confirmed December offer, an inspection or a measured commute. Household ratings are never published, so rankings here leave out the deal-breaker and preference scores that the private app uses. Units marked as deal breakers are removed before publishing.