ani-web/server/providers/animekai.py

"""providers/animekai.py — animekai.be source (hianime-family CMS, own ID space).

Each episode lists one or more servers, allocated per-episode across TWO
backends, handled by separate _backend_* functions (add a third backend =
add one branch in _master_m3u8):

- my.1anime.site  -> direct progressive MP4 (<source> on the embed page).
  Independent CDN; its file host demands a Referer (403 without).
- megaplay.buzz   -> MegaPlay player, shared with hianime via
  upstream.megaplay_master. Same CDN family.

master_m3u8() prefers 1anime first (independent of the MegaPlay CDN), then
megaplay. The returned "master" is just "the URL the player fetches":
.m3u8 -> hls.js, anything else -> native <video> through the /stream
proxy's Range path (app.py decides by suffix — providers stay agnostic).

Politeness: every endpoint rides the shared _cached TTL cache (search
300s, details 3600s, episodes 300s — airing shows gain eps, sources
3600s, master 3600s), so repeated UI requests cost the provider one fetch.
"""
import re
from urllib.parse import urlparse

from ..upstream import _cached, _get, megaplay_master

BASE = "https://animekai.be"
RETRIES = {"sub": 300, "details": 3600, "episodes": 300, "sources": 3600, "master": 3600}

# ponytail: process-local slug<->id registry. details(slug) always precedes
# episodes(anime_id) in the app flow (search -> details -> episodes), so an
# id reaching episodes() was minted by this process's details(). A restart
# orphans in-flight ids (user re-navigates) — same failure mode as any
# source switch.
_REG, _BYID = {}, {}
_N = 0

# host -> Referer its file CDN demands; {} entry = allowed, no headers.
# stream_meta() non-None is the /stream proxy's allowlist.
STREAM_HOSTS = {}


def _search(q):
    html = _get(f"{BASE}/browse", params={"keyword": q, "sort": "most_viewed"}, timeout=20).text
    out, seen = [], set()
    for m in re.finditer(
            r'<a class="title" href="[^"]*/watch/([^"?/]+)"[^>]*title="([^"]*)"', html):
        slug, name = m.group(1), m.group(2)
        if slug in seen:
            continue
        seen.add(slug)
        img = re.search(r'<img[^>]+src="([^"]+)"', html[max(0, m.start() - 400):m.end() + 200])
        out.append({"id": 0, "slug_id": slug, "name": name,
                    "poster": img.group(1) if img else None})
    return out


def _details(slug):
    t = _get(f"{BASE}/watch/{slug}", timeout=20).text
    global _N
    if slug not in _REG:
        _N += 1
        _REG[slug] = _N
        _BYID[_N] = slug
    title = re.search(r'property="og:title" content="([^"]+)"', t)
    name = re.sub(r'\s*Episode\s*\d+.*$', "", title.group(1)) if title else None
    name = re.sub(r'\s*[-|–]\s*AnimeKai\s*$', "", name or "").strip() or None
    post = re.search(r'property="og:image" content="([^"]+)"', t)
    syn = re.search(r'<div class="desc text-expand"[^>]*>([\s\S]*?)</div>', t)
    synopsis = re.sub(r"<[^>]+>", "", syn.group(1)).strip() if syn else None
    return {"id": _REG[slug], "name": name, "synopsis": synopsis,
            "mal_id": None,
            "poster": post.group(1) if post else None,
            "seasons": []}


def _episodes(anime_id):
    slug = _BYID.get(anime_id)
    if slug is None:
        raise RuntimeError(f"animekai: unknown anime id {anime_id} (stale after restart?)")
    t = _get(f"{BASE}/watch/{slug}", timeout=20).text
    out, seen = [], set()
    for num in re.findall(rf'/watch/{re.escape(slug)}/ep-(\d+)"', t):
        n = int(num)
        if n in seen:
            continue
        seen.add(n)
        out.append({"id": anime_id * 1000 + n, "number": n, "filler": False})
    out.sort(key=lambda e: (isinstance(e["number"], str), e["number"]))
    return out


def _sources(slug, num):
    d = _get(f"{BASE}/watch/{slug}/ep/{num}/sources", timeout=20).json()
    return d.get("sources") or {}


_sources_cached = lambda slug, num: _cached(
    f"animekai:sources:{slug}:{num}", RETRIES["sources"], _sources, slug, num)


def _languages(ep_id):
    anime_id, num = divmod(ep_id, 1000)
    slug = _BYID.get(anime_id)
    if slug is None:
        return []
    s = _sources_cached(slug, num)
    out = []
    if s.get("sub"):
        out.append({"code": "jpn", "name": "Sub", "embed_url": "sub"})
    if s.get("dub"):
        out.append({"code": "eng", "name": "Dub", "embed_url": "dub"})
    return out


def _backend_oneanime(source_url):
    """1anime embed -> direct MP4 url, or None. The file CDN 403s without a
    Referer, so the proof fetch doubles as the referer discovery."""
    page = _get(source_url, headers={"Referer": BASE + "/"}, timeout=15).text
    m = re.search(r'<source src="([^"]+)"', page)
    if not m:
        return None
    url = m.group(1)
    up = urlparse(source_url)
    host = urlparse(url).netloc
    # prove it serves us AND learn the referer the file host demands
    for refr in (f"{up.scheme}://{up.netloc}/", f"https://{host}/", None):
        try:
            _get(url, headers={"Range": "bytes=0-1023", **({"Referer": refr} if refr else {})},
                 timeout=10)
            STREAM_HOSTS[host] = {"Referer": refr} if refr else {}
            return url
        except Exception:
            continue
    return None


def _backend_megaplay(source_url):
    """MegaPlay embed -> master m3u8 url, or None (shared chain)."""
    master, _subs, refr = megaplay_master(source_url, BASE + "/")
    if not master:
        return None
    host = urlparse(master).netloc
    STREAM_HOSTS[host] = {"Referer": refr} if refr else {}
    return master


_BACKENDS = (("1anime", _backend_oneanime), ("megaplay", _backend_megaplay))


def _master_m3u8(ep_id, lang="eng"):
    anime_id, num = divmod(ep_id, 1000)
    slug = _BYID.get(anime_id)
    if slug is None:
        raise RuntimeError(f"animekai: unknown anime id {anime_id} (stale after restart?)")
    s = _sources_cached(slug, num)
    want = "sub" if lang in ("jpn", "sub") else "dub"
    # 1anime first: independent CDN, so it rides through MegaPlay outages
    for mode in ((want, "sub") if want == "dub" else (want,)):
        servers = s.get(mode) or []
        if not servers:
            continue
        for _name, url in ((x.get("server_name"), x.get("source_url")) for x in servers):
            for _bname, backend in _BACKENDS:
                if url and _bname in url:
                    try:
                        out = backend(url)
                    except Exception:
                        out = None
                    if out:
                        return out
    # distinguish "no servers listed" (legit None = no dub) from "listed but
    # dead" (raise -> ANI_SOURCE fallback), same semantics as hianime
    if s.get(want) or (want == "dub" and s.get("sub")):
        raise RuntimeError(f"animekai: ep {num} servers listed but none resolved")
    return None


# Provider interface. Cache keys namespaced "animekai:"; TTLs keep the
# provider from seeing the same request twice (see module docstring).
search = lambda q: _cached(f"animekai:search:{q.lower()}", RETRIES["sub"], _search, q)
details = lambda s: _cached(f"animekai:details:{s}", RETRIES["details"], _details, s)
episodes = lambda i: _cached(f"animekai:episodes:{i}", RETRIES["episodes"], _episodes, i)
languages = lambda e: _cached(f"animekai:languages:{e}", RETRIES["sources"], _languages, e)
master_m3u8 = lambda e, l="eng": _cached(f"animekai:master:{e}:{l}", RETRIES["master"], _master_m3u8, e, l)


def stream_meta(url):
    """{Referer} (or {}) for our own file hosts, else None — non-None
    doubles as the /stream allowlist."""
    return STREAM_HOSTS.get(urlparse(url).netloc)


def subtitles(ep_id):
    # 1anime MP4s carry no separate subs; megaplay subs are in the tracks
    # the player bakes in — nothing external to hand the frontend
    return None