# Changelog of Zillow Scraper: Property Details, Zestimates & ZIP Listings (`humble-echidna/zillow-scraper`) Actor

- **URL**: https://apify.com/humble-echidna/zillow-scraper/changelog.md
- **Full Actor documentation**: https://apify.com/humble-echidna/zillow-scraper.md

## Changelog

Versions follow MAJOR.MINOR.PATCH (`src/version.py`); Apify shows MAJOR.MINOR from `.actor/actor.json`.
Every run logs its version and records it in the `RUN_STATS` key-value record.

### 0.1.0 (2026-11-04)

First release.

- Zillow property pages (`/homedetails/<slug>/<zpid>_zpid/`) by link, ZPID or US street address (Zillow's own
  address autocomplete on www.zillowstatic.com finds the ZPID), one flat result per home: address, coordinates,
  status, price, Zestimate, rent Zestimate, beds, baths, area, lot, year built, home type, HOA, tax rate, days on
  Zillow, last sale, views and saves, description, photos (largest JPEG), MLS id and name, listing agent and
  brokerage with phone, parcel id, Walk/Transit/Bike Score, filled-in listing facts, price history, tax history by
  year, schools.
- Locations: ZIP codes, "City, ST", state codes, or a Zillow place or search link (read for its place name only, never
  fetched), listing every home of a listing type (for sale: by agent, by owner, new construction, auction, with pending
  homes on request; for rent; recently sold) from Zillow's property sitemaps, matched on the `-ST-ZIP` end of each home's link; newest
  sitemap changes first, locations sharing the result limit. Homes whose live status has moved to another type are
  left out and not charged.
- Only homes not returned before (`onlyNew`), remembered per location and listing type in `zillow-scraper-seen` in
  the user's account, written only after a home reached the dataset. Homes found moved to another type aren't opened
  again for 7 days.
- Property pages through Apify's residential proxy (US), one IP per visitor, three visitors at one page a second each;
  a refused visitor (HTTP 403 or PerimeterX's page) is replaced, at most 30 refusals a run. Sitemaps and the address
  lookup through datacenter IPs. Brotli page bodies (27% fewer proxy bytes than gzip), decoded in bounded steps
  through `mms_common.body.register_decoder`.
- Charged per home returned (`apify-default-dataset-item`, plan-tiered), plus the start fee; "Max results per run"
  and the maximum cost per run stop page fetching once covered.
- Failure isolation: a failing line only affects itself; a run where every line fails ends SUCCEEDED with the reason.

Source check (2026-11-04):

- robots.txt (https://www.zillow.com/robots.txt, `*` group, which applies to our token): allows `/homedetails/`
  (except owner views, rental-manager resources and `altId` variants, none of which the actor builds) and the
  `/xml/` sitemaps it lists itself; disallows `/homes/` search pages, `?searchQueryState=`, `/api/`, `/graphql/` and
  `/async-create-search-page-state`, which the actor never requests. www.zillowstatic.com/robots.txt disallows only
  static assets, not `/autocomplete/`. Enforced on every request by mms_common (no robots.txt exception).
- Terms of Use (https://www.zillow.com/corporate/terms-of-use/, read 2026-11-04) §5 forbid "automated queries
  (including screen and database scraping, spiders, robots, crawlers...)", and §4C allows copying "without the aid of
  any automated processes" only; MLS listing data is for consumers' personal, non-commercial use. Stated in the
  README's legal answer.
