Vet fallback stream sources: sportsurge is a single point of failure, and no source lists women's college basketball #86
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Why
SPORTSURGE_BASE(v2.sportsurge.net) is the only wired stream source (SOURCE_BUILDERSinbackend/app/config.py). Every sport's streams flow through it, and it goes down from time to time. A vetted fallback source would de-risk the whole app.ff1a74f) correctly makes those games "not listed" instead of mis-matching them to men's rows. If a source exists that lists WCB, this is also the ticket to wire it.Verified 2026-11-03 with the app's own tooling
Fetched the hub through the LAN FlareSolverr (3.x,
CLOUDFLARE_SOLVER_URL), parsed withresolver/hub.py's own parser, matched against ESPN's scoreboards (group 50, ESPN's day) with the app's ownmatchup_matches— the exact pipeline the app runs:/watch-ncaab-streams/rows: 29. ESPN men's games: 29 — 29/29 match. ESPN women's games: 40 — 0/40 match any row.watch-ncaab-streams/andwatch-wnba-streams/but no women's-college section at all; the site-wide live list that day (61 rows across all sports) contained zero women's rows.Sites checked live, 2026-11-03 (all via the LAN solver; fetch times ~20:00-20:35 UTC)
/ncaawstreams→ 404)thetvapp.plus — the find (deep dive 2026-11-03 20:15-20:32 UTC)
/v13is what the server renders — 31 "hot" games that day, 5 of them taggedNCAA Women's Division I: Michigan State vs Valparaiso (6:30 PM ET), Indiana vs FGCU (7:00), SMU vs Le Moyne (7:30), Mississippi State Lady Bulldogs vs North Alabama (7:30), Wisconsin vs Bradley (7:30)./watch/ncaaw-streams301 →/v13; sitemap shows 8 league sections, none women's-college.div#my-jwplayerwith a generic iframehttps://gooz.aapmains.net/new-stream-embed/(no per-game token) pluschangeStream(streamId)JS that swaps the iframe tonew-stream-embed/<streamId>; thestream-btn-*buttons render only at game start.gooz.aapmains.net: nginx, not Cloudflare-gated (plain curl works), empty 39-byte doc pre-game; its CSPframe-ancestorswhitelists isportsurge.ws, sportsurge.ws, buffstreams, crackstreams — i.e. the same embed infra family as sportsurge. Promising: if it plays in-game it should be the same JW-player/m3u8 pattern the app already extracts.stream-btnlist, the real embed, and the m3u8. Result to be appended to this ticket.Why WCB is structurally hard (2026-27 rights)
Forum landscape (Reddit / FMHY)
Vet plan (next steps, in order)
stream-btn-*list + realnew-stream-embed/<streamId>, extract the m3u8. Decides whether this is a viable per-game WCB adapter candidate. In flight.docs/stream-sources.md(§3.3) — m3u8 extracted through the app's own resolve path and served through/stream/proxywith the first sub-resource 200. "Alive and lists games" is not "working".docs/stream-sources.md(proposed; that doc is the repo's evidence record and the gateconfig.pypoints at).Risks / constraints
Evidence
/private/var/folders/tq/vsl1bknn6ql_njl_d3j7dxxw0000gn/T/opencode/wcbb-check/and…/subagent-a/(temp dir; volatile — key lines quoted above; will be folded into the docs addendum in step 6).docs/stream-sources.md(2026-10-21 survey + method),backend/app/config.py(SPORTSURGE_HUB_SECTIONS,SPORTSURGE_UNLISTED_FEEDS), commitff1a74f(PR #83).Update 2026-11-03 (~21:15 UTC): broader hub survey + channel players complete. All fetches via the LAN FlareSolverr, 20:19-20:45 UTC. Raw HTML/JSON per site in the temp dir (subagent-b/); summary below.
Per-game hubs — none of the other 13 lists WCB
The DaddyLive family — the standout (dlive.sx → thedaddy.pk / dlhd.st)
Other channel players (the linear route)
(iptv-org sample URLs for the ones it has: ACC Network 23.239.31.26:8989/accnetwork/index.m3u8, ESPN HD 85.237.89.160:9590/usa-s/ESPN-HD/index.m3u8, ESPN 2 / ESPN U / SEC Network on the same hosts — fetched 20:19 UTC; public-IP streams are the least durable of the three.)
Reddit corroboration (search-index snippets; thread JSON is bot-blocked)
Consolidated position (as of 2026-11-03 evening ET)
Across 15 per-game hubs + 4 channel players, only two routes carry WCB at all:
Everything else (including all three ppv.st-platform sites, StreamFree, and both StreamEast domains) lists zero women's games. The Reddit record agrees: no site has persistently carried NCAAW; the named sites (methstreams, vipleague.im, vipbox.lc) are dead or unreliable.
Vet-plan steps 1-2 are now done except the in-game check; next re-check window: The CW's first WCB weekend (2026-11-13/15) — cheap re-verification of the negative list, and the first real test of the CW channel route with a known schedule.
Vet-plan step 1 result — thetvapp.plus in-game check, fired 2026-11-03 23:40:27 UTC (tip-off 23:30 UTC; the check ran as a detached process that survived the session ending).
Verdict so far: this is a genuinely wireable per-game adapter shape — same embed family as sportsurge (the host's frame-ancestors CSP whitelists the sportsurge domains), no solver needed on the embed/playlist hosts, one-hop from page to stream URL. The one unverified link is load-playlist → m3u8 → first-segment 200, which only works on a live game — running that against tonight's live slate now (follow-up comment).
Update 2026-11-04 (~22:30 UTC): the channel route is now mechanically proven end-to-end, live.
Tonight's slate, from the app's own ESPN endpoint (basketball/womens-college-basketball/scoreboard?dates=20261104&groups=50 — works fine; an earlier 400 in this session was my probe using a wrong path): 36 games; 3 live at 22:20Z (Clinton-UNC Asheville 38-18, Erskine-Furman 13-7, Montreat-App State 18-3); 8 more from 22:30-23:30Z. Of all of them, the DaddyLive feed (ntv.cx) maps exactly one to a real channel: Campbell vs NC State, 23:00Z, ACC Network USA (stream-664). Every other game tonight — including the three currently live — is "Channel Not Listed" in that feed (i.e. no free linear path; they are the ESPN+/no-broadcast slate).
The dembed.st player architecture (DaddyLive's current channel delivery), verified against a live stream:
What wiring this would take: a token fetch per playback session with the Origin header, ~2-minute token renewal, pass-through of the (signed, TikTok-CDN) segment URLs, and a new unwrap step in the proxy (PNG pixel decode + gunzip; PIL does it in ~1 s for a 4 MB segment, or the manual filter pass). That unwrap is the one genuinely new piece versus the sportsurge path. This same architecture serves the per-game ESPN+ USA capture channels (yesterday's two WCB games), so one adapter covers both the linear route and the ESPN+ capture route.
Evidence in the temp dir: dembed-664-1104.html (full player source), accn-manifest-1104.txt, accn-seg-1104.png (the decoded segment), ntv-1104b.json (tonight's channel mapping), espn-wcb-appurl-1104.json (tonight's slate).
Update 2026-11-04 (~22:45 UTC): decision made, repo work started.
Route A pin — done on a live CFB game (2026-11-04 23:34Z, UMass vs Ball State, live since 23:00Z). No 8 PM wait needed: the 7/7:30 PM ET slate (CFB + NHL) was live with per-game embeds while the 10 PM slate still carried the generic embed — the pre-game/in-game distinction re-verified on the same page read.
The full chain, every hop plain HTTP with the shared User-Agent and the referring page as Referer (no Origin, no Cookie):
Notes: the variant host rotates with master refreshes (malek5 to halombarat2 within ~20 min); the media playlist lists a 3-segment window (~12 s) and a segment that scrolls out 404s — a verify's segment read can hit that window, which is transient; the segments themselves answer 200 with any Referer.
The app's pipeline already consumes the master shape (verify.py and the channel puller both follow a master to its best variant) and the adapter's m3u8-body case takes it unchanged; regression test added (test_the_playlist_shapes, MASTER_BODY).
Hand-test (task 4.2) on the running app, same game: POST /api/games/college-football-401866455/resolve with source thetvapp — first attempt 502 (not verified: segment 404, the transient window above), second verified in 6.9 s, 200; GET /stream/game/college-football-401866455?s=thetvapp — 200, channel tuned, segments listed; first sub-resource media/0.ts — 200, 2.7 MB MPEG-TS; the raw /stream/proxy path too: master rewritten, variant m3u8 200, segment 200 (2.9 MB). The channel ran a recovery round on the thetvapp source mid-test as well.
Recorded in docs/stream-sources.md (2026-11-03/04 section, Route A last hop pinned). Change thetvapp-stream-source is complete: full lane 926 passed / 1 skipped, openspec validate --strict clean. Ready for the PR.
Shipped.
Route B (the DaddyLive dembed channel route) remains the recorded follow-up, re-check windows 2026-11-13/15 (CW), 2026-12-02/03 (SEC/ACC), March 2027 (tournament).