Repository navigation
Engine update, online library, filtering, categories/tags, new sources, mobile UI optimizations - #7
Closed
morpheasgr wants to merge 109 commits into
Closed
morpheasgr wants to merge 109 commits into
morpheasgr wants to merge 109 commits into
Conversation
Browsing:
- Add shared listing parser (sites/listing.rs) that merges every anchor
linking to the same video, so titles come from title/aria-label/alt
instead of quality badges ("1080p"), noscript markup or nothing.
Fixes XVIDEOS/xHamster titles and XNXX/YouPorn cards being dropped.
- RedGifs: browse via the v2 API with a cached temporary token (the
site is a JS app, so HTML scraping never returned results); gif
lookups are now authenticated and return direct MP4 URLs.
- ThisVid: scrape search/tag listings directly (yt-dlp has no search
extractor for it).
- Chaturbate: use the username as the card title and strip tag markup
from room subjects.
- Surface yt-dlp browse errors and Reddit login requirement instead of
silently returning empty pages.
Cookies:
- New in-app login browser (cookie_login.rs): desktop opens a separate
webview window and reads cookies via cookies_for_url; Android uses a
new CookieBrowserPlugin Kotlin overlay (full-screen WebView +
CookieManager). Captured cookies (incl. HttpOnly) are saved as
Netscape through the existing vault path.
- "Log in with browser" button in Settings -> Cookies and on
cookie-required site pages; cookie site picker lists all sites.
- Vault: treat #HttpOnly_ lines as cookies, not comments.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The APK stored .so files uncompressed with extractNativeLibs=false (AGP default), so YoutubeDL.init / FFmpeg.init could not unpack libpython.zip.so / libffmpeg.zip.so and every yt-dlp/ffmpeg call failed with "not initialized". - patch-android-ytdlp (.ps1/.sh): set jniLibs.useLegacyPackaging = true - validate-android-build: fail when it is missing - YtDlpPlugin: include the init exception in "not initialized" errors - Settings hint + mobile-android troubleshooting entry Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…n init With native libs now extracted, YoutubeDL.init / FFmpeg.init get past the early failure and run commons-compress / commons-io / Jackson code that R8 had renamed in release builds, crashing the app at launch. - proguard overlay + patch scripts: keep org.apache.commons.compress, org.apache.commons.io and com.fasterxml.jackson (added to existing projects too) - YtDlpPlugin: catch Throwable (LinkageError etc.) and give the plugin scope a CoroutineExceptionHandler so init failures are reported instead of killing the process Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Backend
- BrowseQuery.filters (sort / period / duration / quality option ids)
threaded through the browse command, LAN API query params and the
browse cache key; SiteInfo.browse_options advertises what each site
supports. Unknown ids are dropped before they reach a URL.
- SiteAdapter::browse_options / list_categories, list_site_categories
command + GET /api/sites/{id}/categories (6 h in-memory cache).
- Shared category-index parser (sites/listing.rs) and with_query helper.
- Per-site support, verified against live filter menus:
PornHub (o/t/min_duration/max_duration/hd), XVIDEOS (sort/datef/durf/
quality), XNXX (path segments), xHamster (search query params,
category path sorts), YouPorn (/search/<sort>/), Eporner (API order),
ThotHub/ThisVid (KVS sort_by), Reddit (hot/new/top/rising + t),
RedGifs (order). Categories for PornHub, XVIDEOS, XNXX, xHamster,
YouPorn, Eporner, ThisVid and Chaturbate (top hashtags).
- Fixes: ThotHub page 2+ repeated page 1 (from_videos / from paging),
XNXX pages were off by one, xHamster category paging, PornHub
category names were scraped as video counts (now parsed from the
embedded JSON catalog).
Frontend
- BrowseFilterBar: site sort/filter selects (reload from page 1) plus
on-device text (-exclude) and length filters over loaded results.
- SiteCategoryBrowser on site hubs; useSiteInfo hook for live options.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The site hub (/browse/$site) holding the category browser had no entry point. Selecting a site chip now shows a "<Site> categories" button. siteHasCategories() falls back to a static id list when the backend site list (with browse_options) is unavailable. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
R8 stripped YtDlpPlugin$Args / $MediaArgs / $KeepAliveArgs / $ResumeArgs to empty shells (no fields, no constructors), so Tauri's Jackson parseArgs failed with "Cannot construct instance of YtDlpPlugin$Args (no Creators ...)" and every on-device download and stream resolve broke. Keep com.archhive.app.**Args, and add the rule to already-patched projects in patch-android-ytdlp.ps1. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- PornHub retired /categories/<slug> (404 -> "browse failed"). Resolve slugs to numeric c= ids from the live catalog (straight / gay / lesbian, 12 h cache; gay slugs carry a -gay suffix) and fall back to search. Trans uses the main index: /transgender/video* now 404s. - PornHub listings scope to the result grid (#videoSearchResult, #videoCategory, ul.search-video-thumbs) so header "hottest" / recommended dropdowns no longer lead results; skip playlist "Play All" links (pkey=). - PornhubCategoryBrowser is now a picker: a category opens the browse page (sort / filters / infinite scroll) with ?orientation= in the URL; the browse cache key includes orientation. - Browse page loads page 1 on open instead of waiting for "Browse". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Phones never host LAN: ensure_lan_server / regenerate_lan_server refuse on mobile (they would bind 0.0.0.0). The on-device loopback server reports "loopback": true / "lan": false in /api/health and no longer advertises the phone's Wi-Fi IP; the status chip says "(on-device)" instead of "(open)". - sites::throttle: >=1 s between requests to the same host, and after a 429 a cooldown (Retry-After or 60 s) that fails fast with a clear "rate limiting" error instead of hammering the site. - Trending: 10-minute session cache shared by /browse and /feed with in-flight de-duplication (Feed scraped every trending site twice per visit); pull-to-refresh clears it. - PornHub playback tries yt-dlp first (what works for downloads on-device), then the page extractor on mobile; both via the Referer loopback proxy. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- IdentifyDialog / EnrichDialog / performer StashDB actions showed a generic "Query failed": Tauri rejects with a string, not an Error, so the backend message was dropped. Fall back to String(e). - StashDB answers a missing/invalid ApiKey with HTTP 200 + GraphQL "not authorized", not 401; map it to a clear API-key message. - Select Scene.release_date (Scene.date is deprecated on StashDB) and retry with `date` on boxes that do not know release_date. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…essage Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- sceneMediaUrl / sceneThumbUrl / fileStreamUrl no longer require settings.lan_enabled on the phone: the 127.0.0.1 loopback server always runs there, but the flag can still read false while it starts (e.g. first launch after reinstall), leaving library files with no playable URL. - Scene player shows why playback failed (probes the stream: missing file, outside the library folder, server unreachable, unsupported format) with an "Open with system player" button, instead of a silent black box. - SceneContextMenu: bottom sheet on phones (large targets, padded above the system navigation bar); the desktop popup is clamped inside the window. "Open in Explorer" is hidden on phones. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…fe rename - Release APKs shipped usesCleartextTraffic=false: patch-android-lan.ps1 tested for any "true" (only the debug build type sets it) instead of a remaining "false". The WebView could not reach the 127.0.0.1 media server, so library playback, thumbnails and the PornHub Referer proxy all failed. The validator now fails on a remaining "false". - "Open with default app" / "Open with system player" on Android go through a new YtDlpPlugin.openFile (ACTION_VIEW + FileProvider content:// URI); the opener plugin's open_path did nothing there. FileProvider roots overlay adds app-private dirs. - Rename file to match title: error when the file is missing, never overwrite an existing file, move the <stem>.jpg thumbnail sidecar too. - Library rename / regenerate thumbnail / open actions show their result or error as a toast instead of failing silently. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Tag/category indexes are no longer capped (XVIDEOS ~2,000 tags, XNXX 2,000, xHamster 655 were cut to the 300-400 most popular, hiding most rare / late-alphabet tags). - Combined tags: a Tag/Category slug of `a,b` intersects tags where the site supports it, verified live: XVIDEOS /tags/a+b (up to 5; tag pages ignore sort/filters, so filtered queries use the word search) and PornHub /video/incategories/a/b (exactly 2, straight only). Other sites and larger selections run as a word search; BrowseOptions.combine_tags advertises the exact limit. - Tag browser: A–Z / most-videos order and a Combine mode (select tags, "Browse N together"), also in the PornHub category picker. - Browse page: tag autocomplete from the site's tag index, and a note when a combination falls back to word search. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
XVIDEOS' /tags page only lists the overall top ~2,000 tags; the full
index is split by first character (/tags/09, /tags/a … /tags/z, up to
2,000 each, ~50k tags). BrowseOptions.category_groups exposes those
sections; SiteAdapter::list_category_group fetches one on demand, and
list_site_categories(site, group) caches it in memory (6 h) and on disk
(<data>/tag-index, 7 days).
- Tag browser: "Popular · 0-9 · A … Z" tabs.
- Browse page autocomplete loads only the group for the tag being typed
and suggests the 50 most-used prefix matches, including after a comma
for combined tags ("milf, zo" -> "milf,zombie-sex").
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
XVIDEOS has ~38 categories (/c/<Name>-<id>, listed on the home page) separate from its tags. They are now a "Categories" group in the tag browser. Category pages take sort / length / quality as path segments (/c/s:uploaddate/d:10-20min/q:hd/Femdom-235) and page as /<n-1>; the filter bar applies to them. Verified live that categories cannot be combined with tags or searches in one URL (c= and tag= are ignored, /c/X/tags/Y 404s), so Combine is hidden on the Categories tab and a combined category slug falls back to a word search on the names. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Eporner's HTML listings take category filters in the URL, verified live: /tag/<query>/[<page>/][<sort>/]?include=a,b&exclude=c (page must come before the sort segment; the reverse redirects to page 1), also on /cat/<slug>/. The API used for plain browsing has no such filter, so requests with include/exclude use the HTML listing (new parser; slugs validated to [a-z0-9-]). - BrowseFilters gains include / exclude (comma-separated category slugs; part of the cache key and LAN query params); BrowseOptions.include_exclude advertises support. - Browse page: "Categories" panel for such sites — tap a category to require it, again to exclude, again to clear; taps are committed after a short pause so several taps trigger one reload. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
/cats/ lists the 86 categories plus the top 500 tags; /cats/A/ … /cats/Z/
list 500 tags each (no further paging), almost none of them on /cats/.
They are now category groups ("Tags", A … Z) in the tag browser and feed
autocomplete; the main tab is labelled "Categories" for Eporner.
Tags open their exact /tag/<slug>/ page (HTML listing, with sort, paging
and category include/exclude) instead of an API word search. Tag slugs
are case-sensitive (czech-ouQ9EM84 works, lowercased 404s), so they are
not lowercased. Verified that ?include= only accepts categories, not tags.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The categories page only lists 655 categories and ~200 tags; its page
data links an A–Z tag index at /categories-0, /categories-a … -z (up to
2,000 tags each, no further paging), ~30k tags in total. These are now
category groups ("0-9", A … Z) in the tag browser and feed autocomplete.
Single tags open their /tags/<slug> listing (sort/HD as path segments
and /<page>, like categories) instead of a word search; combined tags
search with xHamster's own '+' form.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
XNXX tags are search terms (/tags/<t> redirects to /search/<t>), but it
has the same per-letter index as XVIDEOS (/tags/09, /tags/a … /tags/z,
up to 2,000 each). Added as category groups ("0-9", A … Z) with their
own parser; tag browsing keeps the hyphenated /search/<tag> form XNXX's
own tag links use (it ranks differently from /search/a+b).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
docs/tag-combining.md records the live-verified URL schemes: XVIDEOS tag intersection (/tags/a+b, up to 5, no filters), PornHub two-category intersection (/video/incategories/a/b, straight only), Eporner search/tag + category include/exclude, and the sites that only support word search (xHamster, XNXX, others), with how the app maps each one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ThisVid: one category + one tag is a strict intersection (/categories/X/?tag=Y or /tags/Y/?category=X; verified on result pages), only one tag, search + category is not strict. YouPorn: no strict combination; porntags are search aliases. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
/categories/<c>/?tag=<t> is a strict category ∩ tag listing (verified on result pages); it pages and sorts. BrowseFilters.tag carries the tag (validated to [a-z0-9-]; part of the cache key and LAN params) and BrowseOptions.category_tag_filter enables a "Narrow by tag" box on category pages, with autocomplete from ThisVid's tag index (/tags/?section=a … z, now also A–Z tabs in its tag browser). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Checked against result videos' own tag lists: XVIDEOS tag pages are a keyword search over titles and tags. /tags/petite-brunette lists more videos than petite or brunette alone and none of the sampled results carried that tag; /tags/milf+anal and /tags/a+b+c results were missing tags. combine_tags is now 0 for XVIDEOS (combined tags run /?k=...), the /tags/a+b URL path is removed, and the docs are corrected. Also re-verified the remaining strict claims on result pages: PornHub /video/incategories/a/b (10/10) and Eporner include/exclude (13/13). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PornHub searches take one required category (filter_category=<id>) and up to 10 excluded ones (exclude_category=<id>-<id>, dash-separated as the site's own UI sends them); category listings take only the excludes. Verified strict on result video pages. BrowseOptions gains max_include / max_exclude / exclude_only_kinds so the Categories panel enforces the per-site limits; docs updated. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The picker showed a 55-entry built-in list unless "Refresh counts" was tapped; it now loads PornHub's live catalog per orientation (111 straight, 53 gay, 73 lesbian), cached for the session, with the built-in list only as an offline fallback. Trans has no index of its own (/transgender/categories is a 404) and trans categories used to open the straight /video?c= listing. The Trans tab now lists the straight catalog, each entry browsing <category> ∩ Transgender (/video/incategories/<x>/transgender), with "All trans" (c=83) on top. Gay pairs intersect too (/gay/video/incategories/a/b). Category pairs without a page (404) fall back to a word search. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The categories page also links 17 spoken-language listings (/language/<name>, same grid, paging and sort as categories). The picker's Straight tab shows them as a Languages row; `language-<name>` slugs browse /language/<name>. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Saved searches store the listing's sort / filters (new saved_searches.filters
JSON column, MIGRATION_020); checks run the saved listing instead of the
site defaults, and a filtered page is a different saved search from the
unfiltered one.
- Opening a saved search passes its orientation and filters (browse route
search params now carry the filter keys) and no longer pre-encodes the
slug, which double-encoded combined slugs ("milf,anal").
- Save / remove failures show a toast instead of failing silently; browse
errors show Tauri's string rejections instead of "Browse failed".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Replace <datalist> autocomplete (browse tag box, ThisVid "Narrow by tag") with SuggestInput: an in-page list anchored under the input, ~6 rows then scrolling, closed on pick / blur / Escape. Android WebView drew datalist as a native popup that could cover the screen and overlap the keyboard. - Replace the sideways-scrolling site chips and the separate site dropdown with SitePicker: one large button opening a grid of big site tiles (bottom sheet on phones, dialog on desktop) with search, Auto-detect and a "Needs cookies" hint. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bundle each site's best published icon (apple-touch / largest favicon, normalised to 128×128 PNG, 144 KB total) in public/site-icons and show it in the site picker. ThotHub and ThisVid only publish 16–32px icons, so they and plugin sites keep the coloured initial badge. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PornHub serves its mobile layout when a phone-browser login left a platform=mobile cookie. Its tags are the pill row above the title, which the scraper didn't read, so Fetch tags fell back to yt-dlp and got only the channel. Categories and tags are now read by their data-label in either layout, and the mobile uploader link is read too. A video page answering 404 / 410 now means the video is gone: Fetch tags removes it from the Online library and every saved search (it won't come back on Update) and says how many were removed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fetch tags only read videos never fetched before, so videos fetched while PornHub's mobile layout went unread kept just their channel. Select mode gets "Fetch tags N" for the selected videos, and "Re-fetch all shown" fetches every video matching the current filters (all pages, after a confirmation with a rough duration). Both re-read videos done before, replace their tags, can be stopped, and drop videos the site removed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SceneCard ignores taps in selection mode (pages own selection, as the Library Scenes page does with a wrapper), and the Online library had no wrapper, so only "Select all loaded" could select anything. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PornHub answers taken-down videos with a page whose player is replaced
by a removed-video notice ("This video has been disabled"), with HTTP
200 or 403. Fetch tags now removes those like 404s. Other 403s (bot
checks, rate limits, region blocks) still count as temporary failures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fetch tags already removed videos the site took down; tapping one to play it only showed the stream error (HTTP 403 for PornHub's "This video has been disabled" page). When resolving a stream fails with a 403 / 404 / 410-style error, one page load now checks whether the video is gone; if so it leaves the Online library and saved searches, the player says so, and the Online list drops it. The Info dialog does the same. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With a platform=mobile cookie (a phone browser login), PornHub's disabled-video page shows "This video has been disabled." in a .removedVideoContent box instead of the desktop .video-notice, so the phone never saw these videos as removed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A site in maintenance (or misconfigured) can answer 404 for every page, and Info, playback or a "Re-fetch all shown" run would then have dropped all of its videos. A bare 404 / 410 now counts as removed only when the site's home page loads and isn't a maintenance notice (remembered for 5 minutes per site). The sites' own removed-video notices still count directly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
quinn-proto 0.11.16 is flagged by five RustSec advisories published in August and September 2026 (memory-exhaustion and remote-panic issues, fixed in 0.11.17 / 0.11.18), which fails the cargo audit CI step on this branch and on main. Lockfile-only update; cargo audit now reports no vulnerabilities (only the 7 allowed unmaintained/unsound warnings). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A site having trouble could make a whole batch look removed. Fetch tags now only reports removed videos (fetch_online_details no longer deletes them); the Online library page removes them, but first asks when more than half of a batch (and at least 5 videos) came back removed. Cancel keeps them and stops the run; OK removes them and doesn't ask again for the rest of the run. The Info dialog and player still remove a single removed video directly. Also drops a needless `mut` left in the xPomelo adapter. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
xPomelo lists PornZog thumbnails on the main domain (pornzog.com/media/videos/tmb/...), which answers 404; PornZog serves them from tn1.pornzog.com. xPomelo cards now get the image-host URL, and Online library entries saved earlier are mapped when read. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Browsing no longer puts every result (trending rows included) into the Online library. Results are kept for saved searches (browsing one adds its results to it) and bookmarks; Settings has a toggle to keep everything as before. - The player has a Bookmark button. Bookmarks live in the Online library under "Bookmarks" (a reserved scope, so tags, filters and removal work as for saved searches, and clearing history keeps them). - The player reads tags from the video's page when the listing had none, and the Info dialog now does so whenever tags are missing (it skipped every aggregator result because they carry a channel). Details fetched for a video in the Online library are stored on it. - Trending on Feed and Browse share one row of compact cards that open the player; Feed no longer shows full-width Download buttons. - Library History, performer, tag and collection pages, Home and Feed rows use the same 2-column phone grid as Scenes instead of 3 tiny columns. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Pornwitch serves PornHub thumbnails from cdn.pornwitch.com, which answers 403 unless the Referer is pornwitch.com, so they never showed in the app. Thumbnails on such Referer-gated hosts now load through the app's media proxy (phone loopback, Remote LAN, desktop with LAN on), which sends the right Referer and caches them; this also covers videos already in the Online library. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- External links (source URLs, "Open page" when a video won't play) open in an in-app browser on Android instead of the system browser, so nothing lands in its history. It keeps the WebView's disk cache and cookies, blocks common ad / popunder networks and popup windows, and its toolbar can clear the current site's cookies and storage or share the page. The system browser is only the fallback. - A Share button (Android) in the browse player sends the video's URL to the share sheet (copy, browsers, other apps); in the library player it sends the video file itself (FileProvider) to video players and other apps. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With multiple windows disabled, the WebView loaded every window.open() popunder in the current page, so a click anywhere redirected to an ad. New windows are now accepted and refused in onCreateWindow (counted in the title bar). Script- or click-driven navigation of the page to another site asks "Leave for <host>?" first; server redirects and links within the site load as usual. Blocking ad requests trips sites' anti-adblock checks, so it is now an "Ads: on / off" toolbar toggle, off by default and remembered. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With "Ads: off" on, the in-app browser now also blocks the ~51k ad-server domains from EasyList's `||domain^` rules (site-limited, path, wildcard and element-hiding rules are skipped). The list is downloaded in the background, cached in the app's files as a plain domain list and refreshed every 4 days; the built-in list covers the time before it loads or when the download fails. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A "Query" panel builds a search from groups of alternatives (any term of a group, every group required), an exclusion list and optional "rank higher" terms. Terms are whole words or phrases in titles, tags and performers (punctuation and hyphens ignored, so "ass" no longer matches "glasses"); `term*` matches as a prefix. A "Best match" sort ranks by how many group and boost terms each video has. The query runs in the database over all results, is remembered per saved search / Bookmarks / everything, and also drives "Remove all shown" and "Re-fetch all shown". The component is reusable for the offline library. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A live line above the groups spells out the query in parentheses form, and terms inside a group / the exclusion list are joined with "OR". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Terms match singular and plural forms of their last word (blowjob / blowjobs, tit / tits, pussy / pussies); whole-word matching had missed every plural. - Clear and Save buttons at the top of the builder. Save creates a "filtered view" (name, scope, query and length filter) listed under "Filtered views" in the Online library drop-down; it re-runs live, so updates and full fetches of the underlying saved search show up in it. Saving while a view is open updates it; views can be deleted. Views are kept on the device. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Tags the player reads from a video's page were already stored on the Online library entry (only for videos in the library), but the list kept showing the old card (often just a channel from yt-dlp). The player now hands the details back and the card updates at once. - Un-bookmarking takes the video out of the Online library too, unless a saved search also holds it; the Bookmarks list drops it right away. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Shows only videos with an empty tag list, whether or not they have a channel (yt-dlp often gives just the channel), so they can be fixed in one go with Re-fetch all shown. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PornHub answers 410 with "This page is not available" for videos it won't show in the viewer's region or to their VPN address; the video still exists. The removed-video check counted 410 as deleted, so playing such a video could take it out of the Online library, and the player said it "no longer exists". Only 404 (or a site's own removal notice) counts as removed now, and a 410 says the video is blocked for this connection and suggests another VPN location. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sites (PornHub among them) refuse VPN exit addresses with 403 or 410; the 403 hint blamed only the embedded player / cookies. Both the stream error and the player hint now suggest another VPN location first. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Fetch tags, the Info dialog and the player now follow a Pornwitch
/go?id= link once and move the Online library entry (with its saved
searches and bookmarks) to the real source URL, without Pornwitch's
affiliate parameters. Tags are then read from the source site.
- TXXX-platform sites (TXXX, HClips, HotMovs, UPornia, VXXX,
TubePornClassic, VoyeurHit) and beeg render their pages with
JavaScript, so tags are read from their JSON video records instead:
categories / tags, models, uploader, duration. beeg's brace notation
("Big {Tits}") is cleaned and technical tags (1080p, Index, media
orientation, length) are skipped.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Keeps both sides' settings (online_record_all + recommendation prefs), the in-app browser plugin args next to ShareFileArgs, and upstream's R8-off release build plus package-wide ProGuard keeps. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Scenes get the Online library's query builder: groups of alternatives (AND between groups), exclusions and rank-higher terms, matched as whole words or phrases in titles, tags and performers, best matches first. The query is part of the scene filter, so Save as Smart Collection keeps it; that form now opens pre-filled (it used to drop the filter). Also fixes Load more skipping a page of scenes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Work from the
morpheasgr:forkbranch, developed and tested on an arm64 Android phone(release APK) plus the desktop LAN backend. 108 commits, about 158 files.
The signing / quinn-proto commits from #6 are included but are already in
main, so they add no diff.Engine: phones run their own engine
initializes in release APKs: native libs are extracted (
useLegacyPackaging), andcommons-compress, commons-io and Jackson are kept from R8. Init failures are reported
instead of crashing the app.
engine. Saved "standalone" settings migrate to Local (settings store, device sync, mobile
bootstrap); the enum value stays for old settings.
asks for a desktop host or shows all zeros without one. Remote LAN remains optional.
0.0.0.0); the on-device server is loopback-only and no longer advertises the phone's Wi-Fi
IP. Library playback uses that loopback server even while
lan_enabledstill reads false.from the rendered page (KVS, SpankBang,
<video>, og:video, JSON-LD).--impersonate, then through a verification window whose cookiesare saved per site.
(
RuntimeCapabilities.desktopShell), not on phones.New sources
include/exclude, result counts, per-page completeness checks, request pacing that stays under
Cloudflare's 1015 rate limit, a site-down message, and unsupported sources hidden by default.
/go?id=links are resolved to the real source page.Browsing
from
window.initials; RedGifs on the v2 API.xHamster, Eporner and ThisVid, plus the live PornHub category catalog (trans/gay pairs,
"Language in videos").
docs/tag-combining.mdrecords what each site really combines (verified live).
are reported.
"Full fetch" option (including passes over several sort orders).
Online library (new)
db/online.rs). Recording everybrowsed video is an opt-in setting.
TXXX and beeg use their JSON APIs; PornHub is handled in both its desktop and mobile layouts.
HTTP 410 / 403 are treated as region or VPN blocks, not removals, and a batch where most
videos look removed asks before deleting.
plural matching,
term*prefixes, best-match ordering, Clear / Save, and saved filtered views.Offline library
Playback and network
kept per site.
Cookies, Cloudflare and the in-app browser
impersonation and a verification window.
before navigating to another site, Clear site data, Share, and an opt-in ad blocker using
EasyList domains.
Mobile UI
StashDB
Android build
Testing
cargo test --lib: 231 passedbun run lint,bun run format:check,tsc --noEmit: clean🤖 Generated with Claude Code