Remove donor-site leftovers from docs, templates and comments

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -12,31 +12,28 @@ body:
 
         A few things answer most reports faster than we can:
 
-        - **Everything exit 3?** Check whether the run was **headless** before
-          you check anything else. Akamai refuses a headless browser on this
-          site whatever the exit address — measured HTTP 403 from four
-          residential exits and one datacentre address, against HTTP 200 and
-          the full catalogue from those very same addresses with a real
-          window. `--headful` is the default here; a proxy does not fix it,
-          and `--cdp-endpoint` is the other path measured to work.
-        - **`price` null on some rows?** That is almost certainly correct:
-          every blank price is a lot whose reserve has not been met, and the
-          row carries `reserve_price_set: true` to say so. What is NOT
-          correct is a null price with no reserve flag — that one is worth
-          this issue.
-        - **Zero rows?** You may be on a page that has none: one page past the
-          end of a listing, or a search that matched nothing (that one answers
-          200 with "No results" **and 24 suggested lots**, which this scraper
-          deliberately does not parse). Both are exit 4 and correct answers.
-          See [TROUBLESHOOTING.md](../blob/main/TROUBLESHOOTING.md).
-        - **`rating` and `review_count` empty on every row?** Expected: this
-          site publishes no per-lot rating. The seller's feedback is in
-          `seller_score` and `seller_feedback_count` on a `--mode lot` row.
-        - **The run stopped at page 100?** That is the site's own cap, not a
-          bug. Past it the site returns page 100's own lots under HTTP 200.
-        - **Wrong language?** The locale is a path segment (`/en/`, `/nl/`,
-          `/de/`), not something the exit IP decides. Prices are in euro on
-          all 18 locales.
+        - **Everything exit 3?** That is Expedia's "Bot or Not?" handler
+          (HTTP 429) refusing your exit address, and the scoring is noisy:
+          the same address serves some requests and challenges others
+          minutes apart. Slow down (`--delay 10`), move to another exit
+          (`--proxy` / `--proxy-file`), or wait. The log names the vendor
+          the handler picked, e.g. `blocked (datadome-challenge)`; only a
+          reCAPTCHA pick is ever offered to the solver.
+        - **Exit 6 after page 1?** Look at `stop_reason` in the sidecar.
+          `next_page_throttled` means the page turn (`POST /graphql`) was
+          rate limited while the HTML kept answering 200; `cards_missing`
+          means a page delivered fewer cards than its own counter states.
+        - **Every price changed overnight?** Check `stay_dates`. A search
+          with no dates in the URL prices each property on its own cheapest
+          night; pin `startDate`/`endDate` for comparable prices.
+        - **`image_url` or `amenities` empty on most rows?** Expected: the
+          gallery is not mounted below the fold, and the amenity strip is an
+          A/B variant of the card.
+        - **Zero rows?** See
+          [TROUBLESHOOTING.md](../blob/main/TROUBLESHOOTING.md) first.
+        - **Wrong currency or language?** The storefront is the hostname
+          (`vrbo.com`, `fewo-direkt.de`, `abritel.fr`, `bookabach.co.nz`,
+          `stayz.com.au`), not something the exit IP decides.
 
   - type: dropdown
     id: engine
@@ -84,8 +81,8 @@ body:
       render: shell
       placeholder: |
         python3 playwright_scraper.py \
-          --url "https://www.vrbo.com/en/c/333-watches" \
-          --pages 3 --out grills
+          --url "https://www.vrbo.com/search?destination=Orlando,%20Florida,%20United%20States%20of%20America" \
+          --pages 2 --out orlando
     validations:
       required: true
 
@@ -105,7 +102,7 @@ body:
     attributes:
       label: What you expected instead
       placeholder: >-
-        96 products, as the README says a category page yields.
+        50 properties per page, as the README says a search page holds.
     validations:
       required: true
 

.github/ISSUE_TEMPLATE/site_changed.yml

@@ -9,36 +9,28 @@ body:
         its own report because the fix is different from a code bug:
         something on vrbo.com moved.
 
-        Useful to know before filing, because it narrows the fix a lot — this
-        parser reads **two different sources for two halves of a row**:
-
-        - **The rows** come from the site's own SSR payload,
-          `<script id="__NEXT_DATA__">`, which is in the first response and
-          needs no JavaScript. Titles, ids, urls, images, auction ids and the
-          reserve flag all come from there.
-        - **The money** comes from the rendered card
-          (`.c-lot-card__price` and its siblings), because the payload does
-          not carry it: bids arrive at hydration.
-
-        So "no rows at all" and "rows with no prices" are two different
-        breakages with two different fixes, and `price_source` on the rows
-        tells you which happened: `next_data+dom` is healthy, `next_data`
-        means nothing hydrated.
+        Useful to know before filing, because it narrows the fix a lot:
+        there is **no structured data on a listing page**, so every column
+        comes from the rendered card, anchored on Vrbo's own `data-stid`
+        attributes. The card is `[data-stid="lodging-card-responsive"]`,
+        and the price container has two spellings —
+        `data-stid="product-price-summary"` on vrbo.com and
+        `data-test-id="price-summary"` on the local storefronts — so say
+        which host you ran. `price_source` on the rows says which node was
+        read: `card` / `card-a11y` on a listing, `detail` / `detail-a11y`
+        on a property page.
 
   - type: dropdown
     id: what_broke
     attributes:
       label: What stopped matching
       options:
-        - No rows at all on a page that visibly has lots (the payload moved)
-        - Rows come out with no prices (the card, or its price class, moved)
-        - bid_kind is null on rows that have a price (the status label moved)
-        - A null price with no reserve_price_set flag (the invariant broke)
-        - favorite_count empty on every row (the card's counter moved)
-        - Pagination stops early, repeats a page, or runs past the 100 cap
-        - A lot page returns nothing (--mode lot)
-        - The auctions index returns nothing (--mode auctions)
-        - A no-results search returns 24 suggested lots as if they were real
+        - No rows at all on a page that visibly has properties (the card moved)
+        - Rows come out with no prices (the price container moved)
+        - Prices missing on one storefront only (a new spelling of the container)
+        - Fewer rows than the page's own counter says (cards_missing)
+        - Pagination stops early or repeats a page (the next button moved)
+        - A property page returns nothing (--mode property)
         - Something else
     validations:
       required: true
@@ -48,10 +40,9 @@ body:
     attributes:
       label: URL
       description: >-
-        The exact URL, its locale prefix, and how the browser was reached.
-        Please say whether the run was headful or over --cdp-endpoint: a
-        headless run is refused by the site and is not a parser problem.
-      placeholder: "https://www.vrbo.com/en/c/333-watches (locale: /en/, headful local Chrome, exit: NL residential)"
+        The exact URL, which storefront, how the browser was reached
+        (local headful, headless, or --cdp-endpoint) and from which exit.
+      placeholder: "https://www.vrbo.com/search?destination=... (storefront: vrbo.com, headful local Chrome, exit: BE residential)"
     validations:
       required: true
 
@@ -63,12 +54,11 @@ body:
         Whichever of these you can get. A capture beats a description, and
         `--dump-html` writes one on success as well as on failure.
 
-        - The payload as the page actually serves it:
-          `python3 -c "import json,re;h=open('dump.html').read();d=json.loads(re.search(r'id=.__NEXT_DATA__.[^>]*>(.*?)</script>',h,1).group(1));print(sorted(d['props']['pageProps']))"`
-        - For a price problem: one whole `article.c-lot-card__container` from
-          devtools, and the amount the page shows for that lot.
-        - For a `bid_kind` problem: the exact label the card prints
-          ("Current bid", "Aktuelles Gebot", ...) and the locale.
+        - For a price problem: one whole
+          `[data-stid="lodging-card-responsive"]` card from devtools, and
+          the amount the page shows for that property.
+        - For a missing-rows problem: the counter text
+          (`1 - 50 of 300+`) and `cards_missing` from the sidecar.
       render: text
     validations:
       required: true
@@ -78,5 +68,5 @@ body:
     attributes:
       label: Anything else
       description: >-
-        Whether it reproduces on a second engine, whether it reproduces in a
-        second locale, and when it last worked.
+        Whether it reproduces on a second engine, whether it reproduces on a
+        second storefront, and when it last worked.

.gitignore

@@ -16,8 +16,8 @@
 # every CI job died with FileNotFoundError while the whole suite was green
 # locally. Exactly the class of thing only a clean checkout finds.
 !fixtures_generated.json
-catawiki_products.*
-catawiki_products_*.*
+vrbo_products.*
+vrbo_products_*.*
 
 # --- the live-test harness's working directory ---------------------------
 live_results/

CHANGELOG.md

@@ -36,6 +36,37 @@ that is a fix, and the release notes lead with it.
   logged "Upstream page status success". `status` is now a fallback only
   when it is an integer.
 
+### Docs — leftovers from the repos this one was copied from
+
+Text describing another site as if it were Vrbo, removed or rewritten from
+this repo's own README, TROUBLESHOOTING and parser:
+
+- **`captcha_solver.py`** said no challenge had ever been observed here and
+  that a block was a headless browser refused by Akamai with a 394-byte
+  "Access Denied". Both were another site's. It now says what the README
+  measured: Expedia's "Bot or Not?" handler (HTTP 429) picked DataDome on
+  every recorded refusal, and there is deliberately no DataDome path. The
+  "No DataDome solver" section the docstring pointed at now exists.
+- **`CONTRIBUTING.md`**: the site-change anchors (a different site's
+  `data-testid` grid and tile markers), the list of pinned properties
+  (auction lots, reserve prices, bid counts, a 100-page cap) and the
+  `akamai` marker note (which contradicted this repo's own measurement)
+  are rewritten for Vrbo. Also corrected: the canary note that the browser
+  build is the gate, and a "datacentre gets no response at all" claim.
+- **Issue templates**: the bug-report hints and the site-change template
+  described another site's payload, lots and locales; both now describe
+  Vrbo's cards, storefronts and `stop_reason`s. Example commands use a Vrbo
+  search URL.
+- **`diff_runs.py`**: example filenames, a `sold`/`sold_is_floor` note for a
+  column this repo does not have, and a docstring listing
+  `original_price`/`discount_pct`/`in_stock` and JSON-LD provenance. The
+  docstring now lists the tracked fields and the `stay_changed` bucket.
+- **`.gitignore`** ignored another repo's default output prefix; it now
+  ignores `vrbo_products.*`, the engines' real `--out` default.
+- **`SECURITY.md`** named only Akamai as the protection, said the project
+  has no releases (v0.1.0 exists) and pointed at a README section that does
+  not exist.
+
 ## [0.1.0] — 2026-09-14
 
 First release on this repo family's architecture. The previous contents of