Sreality Palindrom Scraper
Pricing
Pay per usage
Scrape property listings from Sreality.cz - Czech Republic's largest real estate portal. Paste a search URL with filters and get structured JSON of every listing.
Pricing
Pay per usage
Rating
0.0
(0)
Developer
SilentService, s.r.o.
Actor stats
0
Bookmarked
2
Total users
1
Monthly active users
an hour ago
Last modified
Categories
Share
Phase 3 - price history tracking (the feature that closed the biggest single competitive gap, against martas_kristof).
enableHistory boolean (default off, opt-in). When on, the
actor persists a {listingId -> {firstSeenAt, lastPriceCzk, lastSeenAt}}historyStoreId string optional. Name of an Apify Key-Value
Store; lets multiple cron tasks polling different filters share
history of the same property pool. Default uses the actor's own KVS.enableHistory: true)isNew - listing wasn't in state before this run.priceChanged - current price differs from the last
observation. Only true when both sides are real numbers; transitions
to/from priceHidden don't trip the flag.previousPriceCzk - the prior observation when price changed,
null otherwise.daysTracked - days since first observation (>=0).firstSeenAt - ISO timestamp of the first observation, stable
across all subsequent runs._compute_history_fields is a pure function, mutates state in place
but never raises; malformed firstSeenAt falls back to 0 days +
current timestamp.Phase 2 hardening after a Šťoura review (verdict 7/10 -> 10/10 over two rounds). No new features; production-readiness fixes.
_pick_locality previously preferred quarter_cz over municipality_cz
in its category priority, so _resolve_city("Praha") could land on
a Holešovice/Vinohrady quarter instead of all-of-Praha. Worse, for
inputs like Praha 6 Sreality returns quarter_cz with
district_id=0, which our code accepted - producing a filterless,
nationwide scrape instead of erroring out. The resolver now walks
Sreality's own ranking and returns the first match with a real
region_id or district_id; quarter/ward results without IDs raise an
actionable error suggesting URL-paste mode instead, with a URL
hint composed from the user's category/offerType inputs.page/scraped values no longer
crashes the run with a TypeError - falls back to (1, 0, 'fresh')
with a warning log.distance when Sreality didn't compute the route.Retry-After header. Previously
treated as a generic HTTPError and retried with the standard
2/4/8s backoff - that's too aggressive if Sreality returns a
longer cool-down window.api_params log values truncated >48 chars. Defensive against
any future opaque token a user might wire into apiParams._resume_from_state and _parse_retry_after extracted as pure
helpers, unit-tested in isolation. Resume status reported with a
string tag ('resume' / 'fresh' / 'signature_mismatch' / 'malformed'Phase 2 - data quality + UX expansion + crash recovery.
Three new fields turn Palindrom into a structured-input scraper as well as a URL-paste one:
city - free-text Czech city, district or street. Resolved via
Sreality's own suggest API (/api/cs/v2/suggest), so accents and
spelling variants are handled automatically. Verified on probe
queries praha, Praha, plzen, Plzeň.category - enum byty / domy / pozemky / komercni / ostatni.
Used together with city.offerType - enum prodej / pronajem / drazba. Used together
with city.The actor now requires at least one of: searchUrl, apiParams, or
city.
houseNumber - parsed from the street component of localityRaw,
handles trailing letters (12a) and split numbers (3/5).includeDetails mode):
poiTransportDistanceM, poiSchoolDistanceM, poiGroceryDistanceM,
poiDoctorsDistanceM, poiRestaurantDistanceM,
poiLeisureDistanceM. Values are the closest match per category in
meters (straight-line). Sreality already collects these via mapy.cz
/ firmy.cz; we just surface them.resumeFromLastRun input (default off). When enabled, the
actor persists {signature, page, scraped} to its key-value store
after every page push. A subsequent run with the same input
signature picks up where the crashed/aborted run left off. The
signature hashes searchUrl + city + category + offerType + sort +
apiParams, so changing any filter dimension forces a fresh start.
State is cleared on successful completion to avoid resuming past
EOF.Phase 1 quick wins after competitive feature audit (15 Apify Store actors).
errors globally across all search pages, so 5 random transient
failures spread across a 100-page run would abort everything. Errors
now reset on each successful page; a page is retried 3 times before
it is skipped, and the run aborts only if 3 pages in a row fail.pricePerSqm - precomputed priceCzk / areaM2 (round to int),
null when price is hidden or area is unknown. Matches a feature 5
of 15 competitor scrapers expose.agentUrl - direct link to the agency profile on
sreality.cz/adresar/<slug>/<id>, lowercase slug verified against a
live 301 redirect.sort - optional override for search ordering. Enum values
verified against the Sreality JSON API:
-date / +date / -price_norm / +price_norm. Empty (default)
keeps whatever ordering the search URL specifies.apiParams - power-user bypass for the HTML bootstrap. When
provided, the actor calls /api/cs/v2/estates directly with these
snake_case params and skips the 725 KB search-page download.
Useful for cron-driven repeat polls of a known filter.dataset_schema.json (was: 1):
overview, pricing, location, media, agent, compact. Apify Console
now lets users switch perspectives without exporting the full row.pricePerSqm math, hidden-price handling,
no-area handling, _build_agent_url slug/id edge cases).palindrom (slug janmatejka/palindrom).
Title becomes "Sreality Palindrom Scraper". The name is a small joke:
the word "palindrom" is itself not a palindrome.scripts/measure_economics.py.
The script now reads DECODO_USER / DECODO_PASS / DECODO_HOST from
the environment and skips proxy scenarios when they're not set. The
previously-hardcoded credentials must be rotated on the Decodo
dashboard - they appeared in builds 0.1 - 0.5 of the old actor.JSONDecodeError in addition to
httpx.HTTPError. If a proxy ever returns malformed HTML in place of
JSON, the run retries (up to 5 times with exponential backoff) instead
of crashing with an unhandled exception.category_type_cb=3) now produce a working URL.
Sreality uses /detail/drazby/... (plural slug) for the URL while the
offerType field stays drazba (singular Czech word). Two separate
maps handle the difference.This release fixes critical data-integrity bugs and ships substantial performance improvements based on a thorough review.
priceUnit field added. Rental listings carry priceUnit: "za měsíc"
while sales have priceUnit: "". Without this, rental prices look like
total prices in the dataset.subtype now covers all 44 Sreality subcategories, not just the 13
flat dispositions. Houses, plots, commercial properties and "other"
(garage, wine cellar, mobile home, ...) all get a meaningful subtype.subtypeId exposed for users who prefer Sreality's numeric ID.[["panel"], []], now ["panel"].images are 1200x900 (was 400x300
thumbnails); thumbnails preserved in imageThumbnails.pageSize default 60 -> 500. Sreality accepts up to 500 listings
per request; 1000 listings now requires 2 search requests (was 17).detailConcurrency
(default 10). Detail-mode runs are roughly 5x faster.maxResults=0 (unlimited)
prevents runaway billing on multi-million-result searches.scripts/discover_slugs.py re-verifies the URL-slug mapping against
live Sreality data.Initial pilot release. Search URL bootstrapping via __NEXT_DATA__,
basic field extraction, Decodo BYOP support, single Praha-byty smoke
test.