fix: retain Zoko opening and ending metadata through playback

2026-10-25 23:01:38 +01:00
parent d8fb648c90
commit 41aa0abd56
11 changed files with 341 additions and 24 deletions
@@ -81,6 +81,8 @@ The canonical handoff still uses `/stream/ani/<AniList ID>/<episode>/sub`, not t

 Resolution uses bounded ordinary HTTPS with certificate verification, no cookies/credentials, no redirect following, and no browser or advertising/challenge script execution. AniList's returned catalog ID must match the requested identity; invalid mappings and GraphQL errors are rejected. The same metadata response can supply the playback title, with `AniList <ID> · Episode <number> (sub)` used when the title is absent. Only the provider's static `window.__P` Base64/XOR configuration is decoded with the first-party `otaku-embed-v1` codec. Poster, download, chapter, view/telemetry and player-script fields are ignored and never fetched.

+ZokoAnime's optional `skip.intro` / `skip.outro` timestamps feed the existing opening/ending bands and skip controls, including queued episodes and saved-episode reopening. Fractional seconds are preserved. Each interval must have finite, nonnegative, increasing boundaries within the actual player-reported duration; missing or malformed timing is ignored without blocking playback. No Miruro lookup or external timing service is used.
+
 The existing process-local loopback HLS relay normalizes playlists for the native decoder. This provider alone permits HTTPS playlists/VTT on `hls.dramahot.top` and HTTPS media resources on that host or the demonstrated `hls1.drama1.cfd` segment origin (default port only). Every manifest URI is checked before registration and fetching, upstream redirects are refused, and only the fixed public `Referer: https://zokoanime.video/` is supplied to those admitted origins. Upstream URLs/tokens remain in memory and are not progress identities. Native global same-origin policy and TLS verification are unchanged; new CDN hosts fail closed until explicitly reviewed. Subtitle redirects retain the existing same-origin-only restriction.

 Resume, queue deduplication and per-item reports use **`anilist:<ID>:<episode>`**, separately from existing `miruro:` and YouTube identities. Queues use the existing version-1 queue document, with the watch URL above as `url`; progress JSON uses the existing version-2 fields with `media_identity: {"provider":"anilist","media_id":"1","episode":1}`. Old Miruro history is neither migrated nor reinterpreted. Resolved-source files also accept `identity.provider = "anilist"` with canonical decimal IDs; as with Miruro files, the caller owns the explicitly supplied source and headers.