# Changelog of Facebook Pages Scraper (`vortex_data/facebook-pages`) Actor - **URL**: https://apify.com/vortex\_data/facebook-pages/changelog.md - **Full Actor documentation**: https://apify.com/vortex\_data/facebook-pages.md ## Changelog All notable changes to this Actor will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/). *** ### \[0.1.105] - 2026-07-12 #### Changed - **Traffic diagnostics now support pricing decisions** - `OUTPUT.trafficStats` reports request counts, response-body bytes, and network errors by Facebook route and by datacenter/residential transport. The values intentionally do not pretend to be billing bytes; exact proxy charges remain available in the Apify run usage report. - **Pricing is one predictable Page-result event** - effective July 27, 2026, one complete Page row costs $0.0055, regardless of how many selected nested posts, comments, media, reviews, or events it contains. The only additional event is Apify's minimum $0.00005 synthetic Actor start. ### \[0.1.104] - 2026-07-12 #### Changed - **Anonymous Page details are residential-first** - the critical `/about/` request no longer waits for repeated datacenter login redirects before using the transport that recovered every tested Page. Supplemental Page metadata and all content sections remain on datacenter proxy; datacenter becomes the final same-endpoint fallback only if residential is exhausted. Cookie-mode is unchanged. - **Page details now use the browser request profile** - `facebookexternalhit` remains limited to content listing routes where it expands embedded edges. The Page `/about/` route uses the Chrome-shaped request profile to avoid the volume-dependent login redirects seen from crawler-UA About requests. - **Residential proxy is an adaptive Page-details recovery only** - ordinary traffic remains on datacenter proxy. When anonymous Page details exhaust the normal route because Facebook keeps redirecting it to login, only the critical `/about/` request moves to a residential exit. Transparency, long About, profile-photo, cookie-mode, and all content sections stay on the datacenter transport; the normal parser then merges their fields. - **Default Page concurrency is now 8** - a 100-page A/B produced the same completeness as concurrency 12 in slightly less wall time and without the cascading login-response behavior seen when retries increased sustained request pressure. API callers can still explicitly choose 1-16. - **Numeric-ID events are no longer duplicated across future and past** - the shared `profile.php?id=...&sk=events` response is now classified by actual event start time or Facebook's explicit past/upcoming marker. Unknown-time records default to past instead of being falsely advertised as upcoming. - **Failed numeric-ID primary bundles get one combined recovery attempt** - only after a real login/unparseable/network failure on the shared numeric route, the Actor retries the same combined Facebook response with a fresh proxy exit before falling back to separate Posts and Photos requests. Handle URLs do not add this retry because a 100-page A/B showed cascading login responses under sustained load. Valid empty feeds are never retried. - **Bulk-run diagnostics now show where time and recovery go** - OUTPUT separates the core `/about/`, Page Transparency, root-profile, and long About routes, including successful attempts, failures, elapsed time, and how many missing fields each route contributed. `recoveryStats` aggregates primary-bundle outcomes and section retry reasons without storing URLs, cookies, or scraped content. - **Current Facebook Page reviews are parsed again** - the reviews tab now recognizes Facebook's newer recommendation-shaped `Story` records, including positive/negative recommendation, review text, author, URL, timestamp, engagement counts, and recommendation tags. The legacy recommendation shape remains supported. - **Auto-detected section URLs are routed per Page** - mixed `/posts/`, `/reviews/`, `/videos/`, `/reels/`, `/photos/`, and `/events/` inputs no longer merge into one global section set that is fetched for every Page. - **First-run input UX is more task-oriented** - the Apify Input tab now starts with Page URLs, "Scraping scenario", and the main item cap. Advanced filters, per-section limits, comments, performance, and cookies are grouped into optional sections. - **Export format was removed from the Console form** - users no longer have to choose an output mode. The Actor writes the simple one-row-per-Page dataset by default, while the old `outputMode` parameter remains accepted for existing API integrations. - **User-facing mode names are clearer** - the form now says things like "All-in-one Page scraper", "Page info / lead list", "Comments from Page posts" instead of asking users to understand internal package names. - **Input is now framed as concrete scraping scenarios** - the main selector is "Scraping scenario" with actor-style choices such as All-in-one Page scraper, Page info / lead list, Posts scraper, Comments from Page posts, Media scraper, Reviews scraper, and Events scraper. API callers can use `scenario` or `task` as aliases for `dataPackage`. - **Safer first-run defaults** - default content cap changed from `50` to `20` items per section, and public comments now inspect `2` post permalinks with up to `5` comments each by default. Users can still raise all limits for heavier runs. - **High-traffic Apify actor compatibility expanded** - `resultsLimit` now works as an alias for `maxResults`, matching Apify's official Facebook Posts / Comments / Reels input style. - **Official Posts date aliases are supported** - `onlyPostsNewerThan` and `onlyPostsOlderThan` now map to the Actor's posts date filter, and date inputs accept `YYYY-MM-DD`, ISO timestamps, or relative values like `7 days`. - **Flat item rows now include migration-friendly aliases** - post/comment/ media/review/event rows include familiar fields such as `postId`, `commentId`, `time`, `text`, `user`, `likesCount`, and `commentsCount` alongside the Actor's native snake\_case fields. - **URL section routing is now available via `dataPackage=auto`** - inputs like `/posts/`, `/reviews/`, `/videos/`, `/reels/`, and `/events/` now resolve to the matching focused Page scenario instead of fetching every section unnecessarily. Plain page URLs still resolve to full-content mode. - **Focused actor compatibility improved** - API callers can use `type` as an alias for `sections`, matching the common pattern used by posts/reviews/ events-style Facebook actors. - **Posts can now be date-filtered** - `startDate` and `endDate` keep only posts within an inclusive `YYYY-MM-DD` range, and the comments pass now opens permalinks only for posts that remain after that filter. - **Default runs now use the all-in-one full-content scenario** - new inputs default to `scrapeMode=fullContent`, combining page details, posts, photos, videos, reels, reviews, and events in one Actor instead of pushing users through separate competitor-style Actors. `pageDetailsOnly` remains available for high-volume metadata/contact enrichment. - **API callers can now select exact sections** - `sections` / `contentTypes` accepts values like `posts`, `reviews`, `photos`, `videos`, `reels`, or `events`; page details are always included so rows keep stable identity and contact metadata. - **Task-oriented jobs replace raw section thinking in the UI** - users can now choose human-readable jobs backed by `fullContent`, `contentMonitoring`, `posts`, `media`, `reputation`, `events`, or `leadEnrichment`. Older `scrapeMode` inputs still work as aliases. - **Full-content root bundling no longer replaces rich page details** - the root page bundle still seeds posts/photos, but `/about/` remains the source of contact/category/rating fields so speedups do not silently drop metadata. - **Page details now merge competitor parity fields** - Page Transparency and About routes can now fill `confirmed_owner`, `CONFIRMED_OWNER_LABEL`, `about_me`, `whatsapp`, `whatsappUrl`, and `pageAdLibrary.is_business_page_active` when Facebook exposes them. - **Page-details concurrency is higher by default** - metadata-only runs now default to `pageConcurrency=10`; full-content runs keep the safer internal default of `6`. - **Authenticated runs now preserve input identity more carefully** - when logged-in Facebook HTML contains unrelated Page/User nodes, the Actor no longer lets those embedded IDs overwrite the requested numeric profile/page ID or handle. Unverified handle IDs are omitted and marked with `pageIdentityUnverified=true`. - **Authenticated requests are paced per page** - cookie mode now limits each page to a small number of simultaneous section requests, reducing timeout storms from sending one Facebook account through too many proxy exits at once. - **Anonymous section fetches are paced too** - each page now fetches up to 3 sections at once instead of all sections simultaneously, reducing proxy timeout storms on batches with `pageConcurrency=6`. - **Core Facebook requests get a little more time to finish** - the default per-attempt request timeout increased from 15 to 20 seconds, while optional empty-prone sections still use their shorter budget. - **Core section network failures get one targeted recovery pass** - if page details, posts, photos, or videos exhaust the normal network budget before emitting data, the Actor retries that section once with a fresh proxy session instead of immediately downgrading the page to a partial result. - **Network retries no longer stay pinned to one slow proxy exit** - initial Facebook page fetches now let each retry rotate to a fresh Apify Proxy session, so a bad exit IP is skipped sooner instead of consuming the whole retry budget. - **False deleted-handle skips are reduced with tokenless oEmbed** - when the anonymous Graph object probe returns 400 for a handle, the Actor now checks Meta's public `oembed_page` endpoint before skipping. If oEmbed confirms the URL exists, the Actor continues with the normal public HTML scrape instead of incorrectly returning `not_available`. - **Rows with public content but missing page details are marked partial** - when Facebook gates the `/about/` page but public listing sections still expose posts/videos/reels, the row now uses `resultStatus=partial_public` and `dataQuality=Partial` instead of looking like a full page-details scrape. - **Bulk default concurrency increased to 6** - controlled Apify bulk tests on real inputs completed cleanly at six pages in parallel, so new runs finish faster without users needing to tune `pageConcurrency` manually. - **Empty public optional sections use a faster network budget** - reels, reviews, future events, and past events still get fetched, but slow proxy retries no longer let mostly-empty tabs hold up an otherwise healthy page. Network timeouts in those optional tabs are treated as an empty public section instead of turning an otherwise complete page into an OUTPUT failure. - **Proxy tunnel failures get one clean section retry** - if Apify Proxy returns `CONNECT tunnel failed, response 594` after the normal HTTP retry budget, the Actor retries that section once with a fresh proxy session before surfacing a section warning. - **Bulk runs now process pages with bounded concurrency** - the Actor scrapes up to 6 input pages at a time by default, configurable with `pageConcurrency` from 1 to 8. Each page still scrapes all supported sections concurrently, so large batches spend much less wall-clock time waiting on slow Facebook/proxy responses. - **Unavailable pages no longer trigger partial fallback recovery** - when Facebook reports a missing handle or returns no renderable page data, the Actor now emits a `not_available` row plus `OUTPUT.skippedPages` diagnostics instead of spending extra proxy traffic on Ad Library, Graph picture recovery, Wayback, Page Plugin, or other incomplete fallback sources. - **False deleted-handle skips get one cheap public-details rescue** - when Graph and oEmbed both say a handle is gone, but Facebook's own public `/about/` route still renders page details, the Actor now keeps that row and continues scraping the remaining sections. The rescue is single-attempt and short-budgeted, so truly dead handles still return cheap diagnostics. - **Slow proxy exits are rotated away sooner on core sections** - page details, posts, photos, and videos now use an 18-second per-attempt network budget with one in-request transport retry. If a core section still fails, the existing section-level recovery gets a fresh proxy session. This caps the damage from one bad exit IP while preserving recovery for high-value data. - **Optional empty-prone sections use a tighter first budget** - reels, reviews, future events, and past events keep one retry but now use a 10-second per-attempt budget, reducing long tails on sections that commonly return no public data. - **Anonymous root HTML is reused as a primary bundle** - the root page route already carries page details, posts, and a rich photo set, so anonymous runs now parse those from the first page document instead of fetching separate `/about/` and `/photos/` routes when the bundle is complete. - **Numeric page events are fetched once** - `profile.php?id=...` pages use the same Facebook `sk=events` route for upcoming and past event data, so the Actor now parses both arrays from one request instead of loading that URL twice. #### Added - **Legacy Facebook URL normalization** - older public links such as `/p/-/`, `/pages///`, `/people///`, and slug-with-ID URLs are now treated as numeric Page IDs instead of being skipped or misread as the handle `people`. - **Timeout guard and progress checkpoints** - OUTPUT is now written during long runs, not only at the end. When the Apify platform timeout is near, the Actor stops starting new pages with a safety buffer, saves partial results, and lists the remaining inputs in `OUTPUT.notStartedPages` instead of being killed without a useful summary. - **Optional authenticated cookie mode** - users can now paste cookies from their own logged-in Facebook browser session into `facebookCookies` to scrape pages that account is allowed to view. The Actor accepts Cookie header strings, JSON cookie-extension exports, and Netscape cookie-file rows; it validates that `c_user` and `xs` are present before starting. Raw cookie values are never written to logs, dataset rows, or OUTPUT. - **Readable result summaries** - every dataset row now starts with `resultStatus`, `dataQuality`, `resultSummary`, `sourceSummary`, `contentCounts`, and `nextAction` so Apify Overview immediately explains whether the row is full public data, partial public data, authenticated data, or unavailable anonymously. - **Friendly OUTPUT key** - the run summary now includes `status`, `message`, `resultBreakdown`, and a compact `pages[]` list with one readable summary per input URL. - **Typed social-media arrays**: `youtube`, `tiktok`, `twitter`, `linkedin` (in addition to the existing `instagram`) — each shaped as `[{"username", "url"}]`. Parsed from every off-Facebook link surfaced by the info-cards (not just the legacy `screenname` slot), so pages that link multiple social networks now expose every one of them as a typed entry. - **`alternativeSocialMedia`** now picks the most prominent off-Facebook social presence by preference order (Instagram → TikTok → YouTube → Twitter/X → LinkedIn) instead of the bare Instagram URL. - **Stub rows for unavailable pages** — when Facebook refuses to render a page anonymously (renamed handle, audience-restricted with no ads, etc.), the Actor now emits a compact stub row matching the popular community scraper's format: `{"url", "error": "not_available", "errorDescription"}`. Drop-in compatibility with downstream pipelines that expect one row per input URL. The verbose diagnostic for each stub still goes to `OUTPUT.skippedPages` for operators who want to know exactly why. - **Dataset schema fields**: `resultStatus`, `dataQuality`, `accessLevel`, `resultSummary`, `sourceSummary`, `dataSources`, `contentCounts`, `nextAction`, `authenticatedSessionUsed`, `recoveredViaAuthenticatedSession`, `youtube`, `tiktok`, `twitter`, `linkedin`, `url`, `error`, `errorDescription`. #### Changed - **README and Actor schemas** now describe the actual input/output behavior more precisely: optional cookie mode, anonymous pagination limits, profile.php numeric-ID handling, unavailable rows, and authenticated unavailable rows. - **User-supplied Facebook cookies** are scoped to Facebook hosts only and raw cookie values are never written to logs, dataset rows, or OUTPUT. - **`intro`** is now correctly extracted from the trailing prose after `" talking about this. "` in `og:description`. Pages without a physical location (creators, online brands) used to lose this field; now they keep it. - **`websites[]`** is built from EVERY info-card's outbound link, not just the three legacy slots (`address`, `screenname`, `website`), PLUS a full-HTML sweep of every `l.facebook.com/l.php?u=` redirect (the format Facebook uses for the page-header social-icons strip). Previously the Instagram / YouTube / TikTok URLs there were invisible to the parser because they live as raw redirect strings inside inline JSON blocks, not as typed card slots. - **JSON-Unicode unescape** in `_unwrap_redirect` now resolves *every* `\\u00XX` sequence (previously only `\\u00253A` and `\\u00252F` were handled, so YouTube / TikTok handles after `\\u002540` = `@` were silently dropped — `youtube.com/@casinocarisa` came out as `youtube.com/`). #### Changed - Skip reason for gated pages now distinguishes the useful current cases: - **handle not found (renamed or deleted)** — Graph/oEmbed cannot confirm the handle. - **page requires login (age-restricted, country-locked, or audience-limited)** — Graph or oEmbed confirms the page exists, but the main public HTML route does not expose renderable data. - **Facebook returned no public data** — Facebook refused to render content and the Actor could not classify the reason further. - Final run summary lists every skipped page with its short reason. #### Fixed - **Unavailable-page runs no longer fail** — if every input page is audience-restricted / renamed / otherwise hidden from anonymous viewers, the Actor now exits successfully after writing the expected `not_available` stub rows and `OUTPUT.skippedPages` diagnostics. These are handled data outcomes, not infrastructure failures. - **Schema validation** — every dataset field declared with strict types now accepts `null` alongside its primary type, matching what Facebook actually ships (many optional fields are `null` for any given page). - **Log noise** — graph.facebook.com probe responses (the 400 / 403 we use to classify pages) no longer surface as user-visible ERROR lines; they're DEBUG. - **Slow-proxy WARN threshold** raised from 10 s to 25 s — Apify Datacenter US through Facebook routinely sits at 8–20 s end-to-end on busy pages, so the old threshold flooded the log with non-actionable warnings. - **curl\_cffi error strings** trimmed in WARN / ERROR lines — the verbose `"Failed to perform, curl: (28) Operation timed out after 15000 milliseconds with 381283 bytes received. See https://curl.se/libcurl/c/libcurl-errors.html first for more details."` style is reduced to the actionable middle phrase. - **Per-section WARN deferral** — section-level fetch / parse failures are now surfaced only when the page actually produces a dataset row. Pages that get skipped no longer pile up red WARN lines for sections that were never going to succeed; the skip reason itself is enough context. - **Section-list 404s** (e.g. `/reviews/` on a page that has no reviews tab) are now silent — they were always handled gracefully by the section-skipping logic; the underlying `ERROR HTTP 404` line was just visual noise. - **Section "not publicly available" INFO** demoted to DEBUG — the per-page summary line already reports the section as "0 reviews" / "0 events"; the per-section info line repeating the same fact for every gated page was log-spam. #### Removed - **Input `proxyConfiguration`** field — the proxy is now hard-coded to Apify Datacenter US so users have one less thing to misconfigure. The Actor continues to use Apify Proxy internally as before. - **Geo-rotation fallback** (DE / BR / GB exits) — empirically verified not to help on the kinds of pages users run into (audience-restricted pages return the same gate from every exit country we tested). Removed to save the ~26 s it added to gated-page runs. #### Performance - **Pre-flight Graph API probe** for handle-based inputs — when Graph returns `400 "Object with ID ... does not exist"` we skip the full HTML scrape entirely. Saves ~25 s per renamed-/deleted-handle in the input. #### Internal - `HttpClientError` now carries a `status_code` attribute, replacing the fragile "search for the HTTP code in the message string" pattern that previously broke when the message phrasing varied between retry-budget exhaustion and direct 4xx returns. - New `quiet=True` parameter on `FacebookHttpClient.request()` demotes retry / failure log lines to DEBUG for known-best-effort fetches (Page Transparency, profile-photo viewer, Graph API/oEmbed probes, section list URLs). *** ### \[0.1.x] — Initial public release Public Facebook Pages scraper. No login, no cookies — fetches anonymously via curl\_cffi with Chrome TLS impersonation and the `facebookexternalhit/1.1` user agent. Per input URL emits one rich dataset row with: - Page details (title, page ID, handle, categories, intro, address, phone, email, website, hours, rating, follower counts, profile photo, cover photo, page creation date, linked Instagram). - Posts, photos, videos, reels (with timestamps, reactions, comments, shares). - Public reviews. - Upcoming and past events. Uses Apify Datacenter US Proxy with per-request session rotation. Soft-fails on 404 / login-redirect for missing sections. Handles HTML entity decoding for non-ASCII content. Eight section types fetched concurrently per page; pages processed sequentially.