Add /api/match/{event_id}/detail route to main.py (the actual running module)

main.py

@@ -1,1233 +1 @@
-"""VNEWS - FastAPI backend with livescore + xemlaibongda highlights + YouTube FPT shorts"""
-import hashlib, re, time, subprocess, json, os, threading
-import html as html_lib
-from datetime import datetime
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from fastapi import FastAPI, Query, Request
-from fastapi.responses import HTMLResponse, JSONResponse, StreamingResponse, Response
-from fastapi.staticfiles import StaticFiles
-from urllib.parse import unquote, quote, urlencode
-import requests
-from bs4 import BeautifulSoup
-
-app = FastAPI()
-HEADERS = {"User-Agent":"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36","Accept-Language":"vi-VN,vi;q=0.9,en;q=0.8"}
-BONGDA_HEADERS = {"User-Agent":"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36","Accept-Language":"vi-VN,vi;q=0.9","Referer":"https://bongda.com.vn/lich-thi-dau","X-Requested-With":"XMLHttpRequest"}
-BASE_BDP = "https://bongdaplus.vn"
-SPACE_URL = "https://bep40-vnews.hf.space"
-_cache = {}
-_cache_ttl = 300
-_cache_ttl_live = 60
-_cache_ttl_yt = 600  # 10 min: RSS feed is cheap, keeps shorts fresh/live
-SHORTS_FALLBACK = [
-{"id":"Lu_iCQ5YwNM","title":"Công an lập hồ sơ xử lý người phụ nữ chửi bới, tát nam tài xế ô tô ở Hà Nội | #shorts","channel":"baodantri7941"},
-{"id":"CwWvijF8BOA","title":"Chú rể Ninh Bình bật khóc nhận món quà bí mật người cha quá cố gửi 26 năm trước | #shorts","channel":"baodantri7941"},
-{"id":"tvPewsc2ph4","title":"Tính năng ẩn trên iPhone giúp giảm mỏi mắt | #shorts","channel":"baodantri7941"},
-{"id":"b1Nxzv9ixlU","title":"Y án 3 năm tù với nữ tài xế uống 8 lon bia lái xe tông chủ tịch xã tử vong | #shorts","channel":"baodantri7941"},
-{"id":"Xp5eTwAZAis","title":"Người đánh hàng xóm tại chung cư ở Hà Nội bị tuyên hơn 4 tháng tù | #shorts","channel":"baodantri7941"},
-{"id":"Htzvwg6iOBM","title":"Xe điện Audi S6 Sportback e-tron có gì đặc biệt? | #shorts","channel":"baodantri7941"},
-{"id":"iMdFmWvYdlo","title":"Cô gái người Nga yêu thời trang và đất nước Việt Nam | #shorts","channel":"baodantri7941"},
-{"id":"IVaRc6moEv8","title":"Người nông dân Trung Quốc đột quỵ, bệnh viện giúp bán sạch 4 tấn táo | #shorts","channel":"baodantri7941"},
-{"id":"uVxqPxToItU","title":"Công an vào cuộc vụ người phụ nữ chửi bới, hành hung tài xế ô tô ở Hà Nội | #shorts","channel":"baodantri7941"},
-{"id":"VAfgNNgZDRs","title":"Khởi tố 4 đối tượng ném bom xăng vào nhà dân ở Đồng Nai | #shorts","channel":"baodantri7941"},
-{"id":"sBH_-zGh0Xw","title":"Vì sao Times New Roman vẫn nổi tiếng sau hàng chục năm? | #shorts","channel":"baodantri7941"},
-{"id":"woKn5f2bLHM","title":"Quảng Ninh ngập sâu diện rộng sau đợt mưa lớn | #shorts","channel":"baodantri7941"},
-{"id":"bcpgRoxbLPw","title":"Giông lốc quật bay mái tôn ở TP.HCM | #shorts","channel":"baodantri7941"},
-{"id":"ZIIC5osy544","title":"Bé trai Trung Quốc rơi từ tầng 11 vẫn sống sót kỳ diệu | #shorts","channel":"baodantri7941"},
-{"id":"uTMJ49NQpyc","title":"Sau lớp mascot 40kg: Câu chuyện mưu sinh của người trẻ ở TPHCM | #shorts","channel":"baodantri7941"},
-{"id":"7Pd6vZ2Lz1M","title":"Hành động ấm lòng của người đàn ông tham gia tìm kiếm 5 học sinh tử vong ở sông Lô | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"SlHLt_ZyPiE","title":"Xử phạt người đàn ông xóa số điện thoại cứu hộ trên cao tốc Bắc - Nam | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"IUOprcJyYr4","title":"Phụ nữ táo bón có phải do lười ăn rau? | SKĐS #shorts","channel":"baosuckhoedoisongboyte"},
-{"id":"YY8ojFNE-AU","title":"Quái xế tự quay clip nẹt pô, đánh võng đăng TikTok bị xử lý | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"OV7_oGdQGII","title":"Bố cô dâu khóc sụt sùi rồi quẩy cực sung gây bão mạng | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"FoxhFyz2skY","title":"Người đàn ông nước ngoài đập phá ô tô, bẻ cần gạt nước ở Đà Nẵng | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"R1oC_I8dFPU","title":"Thanh niên buông tay lái, đứng trên xe máy khi đổ đèo ở Đắk Lắk | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"U0Ft6ChWAIo","title":"Cô giáo kể phút tháo chạy khỏi xe khách trước khi bị lũ vò nát ở Cao Bằng | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"hH0ANeze_4E","title":"Liên tiếp hàng chục con bò bị sét đánh chết trong ngày mưa dông | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"pXWt0QbAzRQ","title":"Va chạm giao thông, người phụ nữ lăng mạ tài xế ô tô | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"UWWLPY1OYt4","title":"CSGT chặn xe khách khống chế đối tượng cướp dây chuyền tại Gia Lai | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"AxhVTQutsuo","title":"Xuất tinh sớm và những hiểu lầm thường gặp | SKĐS #shorts","channel":"baosuckhoedoisongboyte"},
-{"id":"cNy6FgaNxYM","title":"Cô dâu khóc sưng mắt vì 6 chỉ vàng không cánh mà bay trong ngày cưới | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"IDt_S6q59Ro","title":"Chở bạn gái không đội mũ bảo hiểm, thanh niên đấm CSGT | SKĐS","channel":"baosuckhoedoisongboyte"},
-{"id":"LFxJ9Ik6W0A","title":"Mệnh lệnh từ trái tim: CSGT Hà Nội mở đường đưa bé 5 tháng tuổi đi cấp cứu | SKĐS","channel":"baosuckhoedoisongboyte"}
-]
-for _v in SHORTS_FALLBACK:
-    _v["link"]="https://www.youtube.com/watch?v="+_v["id"]
-    _v["img"]="https://i.ytimg.com/vi/"+_v["id"]+"/hqdefault.jpg"
-    _v["source"]="yt"
-SHORT_STATS_FILE = "/data/short_stats.json" if os.path.isdir("/data") else "/app/short_stats.json"
-_short_lock = threading.Lock()
-def _load_short_db():
-    try:
-        if os.path.exists(SHORT_STATS_FILE):
-            with open(SHORT_STATS_FILE,"r",encoding="utf-8") as f:return json.load(f)
-    except:pass
-    return {}
-def _save_short_db(db):
-    try:
-        os.makedirs(os.path.dirname(SHORT_STATS_FILE),exist_ok=True)
-        tmp=SHORT_STATS_FILE+".tmp"
-        with open(tmp,"w",encoding="utf-8") as f:json.dump(db,f,ensure_ascii=False)
-        os.replace(tmp,SHORT_STATS_FILE)
-    except:pass
-
-def _short_default():return {"views":0,"likes":0,"shares":0,"comments":[]}
-WALL_FILE = "/data/wall_posts.json" if os.path.isdir("/data") else "/app/wall_posts.json"
-def _load_wall():
-    try:
-        if os.path.exists(WALL_FILE):
-            with open(WALL_FILE,"r",encoding="utf-8") as f:return json.load(f)
-    except:pass
-    return []
-def _save_wall(posts):
-    try:
-        os.makedirs(os.path.dirname(WALL_FILE),exist_ok=True)
-        tmp=WALL_FILE+".tmp"
-        with open(tmp,"w",encoding="utf-8") as f:json.dump(posts[:100],f,ensure_ascii=False)
-        os.replace(tmp,WALL_FILE)
-    except:pass
-PRIORITY_LEAGUES = ["Ngoại Hạng Anh","FA Cup","Champions League","LaLiga","Copa del Rey","Serie A","Bundesliga","Ligue 1","V-League"]
-LEAGUE_IDS = {"nha":27110,"laliga":27233,"seriea":27044,"bundesliga":26891,"ligue1":27212}
-HL_LEAGUES = {"premier-league":{"path":"anh/premier-league","name":"Premier League","emoji":"🏴󠁧󠁢󠁥󠁮󠁧󠁿"},"fa-cup":{"path":"anh/fa-cup","name":"FA Cup","emoji":"🏆"},"bundesliga":{"path":"duc/bundesliga","name":"Bundesliga","emoji":"🇩🇪"},"serie-a":{"path":"italy/serie-a","name":"Serie A","emoji":"🇮🇹"},"la-liga":{"path":"tay-ban-nha/la-liga","name":"La Liga","emoji":"🇪🇸"},"champions-league":{"path":"cup-chau-au/uefa-champions-league","name":"Champions League","emoji":"⭐"},"europa-league":{"path":"cup-chau-au/uefa-europa-league","name":"Europa League","emoji":"🟠"},"world-cup":{"path":"the-gioi/world-cup-qualifiers","name":"World Cup 2026","emoji":"🌍"}}
-def _cached(key, fn, ttl=None):
-    now=time.time();t=ttl or _cache_ttl
-    if key in _cache and now-_cache[key]["t"]<t:return _cache[key]["d"]
-    try:data=fn()
-    except:data=_cache.get(key,{}).get("d",[])
-    _cache[key]={"d":data,"t":now};return data
-def _get(url,headers=None):
-    h=headers or HEADERS;r=requests.get(url,headers=h,timeout=15);r.encoding="utf-8"
-    return BeautifulSoup(r.text,"lxml")
-def fetch_bongda_api(endpoint):
-    try:
-        r=requests.get(f"https://bongda.com.vn{endpoint}",headers=BONGDA_HEADERS,timeout=10)
-        if r.status_code==200:
-            data=r.json()
-            if data.get("status")=="success":return data.get("html","")
-        return ""
-    except:return ""
-def _parse_match_from_li(li, status_type="live"):
-    match_div=li.select_one("div.match")
-    if not match_div:return None
-    home_el=match_div.select_one(".home-team .name");away_el=match_div.select_one(".away-team .name")
-    if not home_el or not away_el:return None
-    status_el=match_div.select_one(".status a");league_el=li.find_previous("strong");time_el=match_div.select_one(".match-time")
-    home_logo=match_div.select_one(".home-team .logo img");away_logo=match_div.select_one(".away-team .logo img")
-    event_id=""
-    if status_el:
-        href=status_el.get("href","");m=re.search(r'/tran-dau/(\d+)/',href)
-        if m:event_id=m.group(1)
-    spans=status_el.find_all("span") if status_el else [];score="";minute=""
-    if len(spans)>=3:score=f"{spans[0].get_text(strip=True)} - {spans[2].get_text(strip=True)}"
-    if len(spans)>=4:minute=spans[3].get_text(strip=True)
-    if not score and status_el and status_el.select_one(".vs"):score="VS"
-    league=league_el.get_text(strip=True) if league_el else ""
-    return{"home":home_el.get_text(strip=True),"away":away_el.get_text(strip=True),"score":score or"VS","minute":minute,"league":league,"time":time_el.get_text(strip=True) if time_el else "","event_id":event_id,"home_logo":home_logo.get("src","") if home_logo else "","away_logo":away_logo.get("src","") if away_logo else "","status":status_type}
-
-# ===== VIDEO PROXY =====
-@app.get("/api/proxy/m3u8")
-def proxy_m3u8(url: str = Query(...)):
-    try:
-        r = requests.get(url, headers=HEADERS, timeout=15)
-        if r.status_code != 200:return Response(status_code=502, content="upstream error")
-        lines = r.text.strip().split('\n');rewritten = []
-        for line in lines:
-            if line.startswith('#') or not line.strip():rewritten.append(line)
-            else:rewritten.append("/api/proxy/seg?url=" + quote(line.strip(), safe=""))
-        return Response(content='\n'.join(rewritten).encode('utf-8'),media_type="application/vnd.apple.mpegurl",headers={"Access-Control-Allow-Origin":"*","Cache-Control":"public, max-age=300"})
-    except:return Response(status_code=502, content="proxy error")
-
-@app.get("/api/proxy/seg")
-def proxy_segment(url: str = Query(...)):
-    try:
-        r = requests.get(url, headers=HEADERS, timeout=30)
-        if r.status_code != 200:return Response(status_code=502, content="upstream error")
-        data = r.content
-        if len(data) > 188 and data[0:4] == b'\x89PNG' and data[188] == 0x47:data = data[188:]
-        return Response(content=data,media_type="video/mp2t",headers={"Access-Control-Allow-Origin":"*","Cache-Control":"public, max-age=3600"})
-    except:return Response(status_code=502, content="proxy error")
-
-@app.get("/api/proxy/video")
-def proxy_video(url: str = Query(...), request: Request = None):
-    try:
-        req_headers = dict(HEADERS)
-        if request and request.headers.get("range"):req_headers["Range"] = request.headers["range"]
-        r = requests.get(url, headers=req_headers, timeout=30, stream=True)
-        resp_headers = {"Access-Control-Allow-Origin":"*","Accept-Ranges":"bytes","Content-Type":r.headers.get("Content-Type","video/mp4")}
-        if "Content-Range" in r.headers:resp_headers["Content-Range"] = r.headers["Content-Range"]
-        if "Content-Length" in r.headers:resp_headers["Content-Length"] = r.headers["Content-Length"]
-        return StreamingResponse(r.iter_content(chunk_size=256*1024),status_code=r.status_code,headers=resp_headers)
-    except:return Response(status_code=502, content="proxy error")
-
-@app.get("/api/proxy/img")
-def proxy_img(url: str = Query(...)):
-    """Proxy images from sources that block hotlinking (DanTri CDN)."""
-    try:
-        r = requests.get(url, headers={**HEADERS, "Referer": "https://dantri.com.vn/"}, timeout=10)
-        if r.status_code != 200:return Response(status_code=502)
-        ct = r.headers.get("Content-Type", "image/jpeg")
-        return Response(content=r.content, media_type=ct, headers={"Cache-Control": "public, max-age=86400", "Access-Control-Allow-Origin": "*"})
-    except:return Response(status_code=502)
-
-# ===== XEMLAIBONGDA HIGHLIGHTS =====
-def _scrape_xemlaibongda_page(page_path, limit=20):
-    try:
-        url = f"https://xemlaibongda.top/{page_path}" if page_path else "https://xemlaibongda.top/"
-        r=requests.get(url,headers=HEADERS,timeout=15)
-        if r.status_code!=200:return[]
-        r.encoding="utf-8";soup=BeautifulSoup(r.text,"lxml");videos=[];seen=set()
-        for a in soup.find_all("a",href=True):
-            href=a.get("href","")
-            if"/video/" not in href:continue
-            if not href.startswith("http"):href="https://xemlaibongda.top"+href
-            if href in seen:continue
-            seen.add(href);slug=href.split("/video/")[-1].rstrip("/")
-            title=slug.replace("-"," ").title()
-            title=re.sub(r'\d{4}\s*\d{2}\s*\d{2}$','',title).strip()
-            title=re.sub(r'\s+V\s+',' vs ',title);title=re.sub(r'\s+Vs\s+',' vs ',title)
-            img=a.find("img") or (a.parent.find("img") if a.parent else None)
-            img_src=""
-            if img:img_src=img.get("data-src","") or img.get("src","") or img.get("data-lazy","")
-            if not img_src:img_src=f"https://img.refooty.com/thumbnail/{slug}.webp"
-            videos.append({"title":title,"link":href,"img":img_src,"source":"xemlaibongda"})
-            if len(videos)>=limit:break
-        return videos
-    except:return[]
-
-def scrape_xemlaibongda():return _scrape_xemlaibongda_page("",20)
-def scrape_highlights_by_league(league_key):
-    if league_key not in HL_LEAGUES:return[]
-    return _scrape_xemlaibongda_page(HL_LEAGUES[league_key]["path"],20)
-
-def scrape_all_league_highlights():
-    results = {}
-    def _fetch(key):return key, scrape_highlights_by_league(key)
-    with ThreadPoolExecutor(8) as ex:
-        futs = [ex.submit(_fetch, k) for k in HL_LEAGUES]
-        for f in as_completed(futs):
-            try:
-                key, vids = f.result()
-                if vids:results[key] = vids
-            except:pass
-    return results
-
-def extract_xemlaibongda_video(url):
-    try:
-        r=requests.get(url,headers=HEADERS,timeout=15)
-        if r.status_code!=200:return None
-        r.encoding="utf-8";soup=BeautifulSoup(r.text,"lxml");video=soup.find("video")
-        if video:
-            src=video.get("src","");poster=video.get("poster","")
-            if not src:
-                source=video.find("source")
-                if source:src=source.get("src","")
-            if src:return{"src":src,"poster":poster,"type":"hls" if".m3u8" in src else"video"}
-        m3u8s=re.findall(r'(https?://[^\s"\'<>]+\.m3u8)',r.text)
-        if m3u8s:
-            og=soup.find("meta",property="og:image");poster=og.get("content","") if og else ""
-            return{"src":m3u8s[0],"poster":poster,"type":"hls"}
-        return None
-    except:return None
-
-# ===== YOUTUBE SHORTS =====
-# Permanent channel IDs (resolved from the @handles) used for the stable RSS feed.
-_YT_CHANNEL_IDS={
-    "baodantri7941":"UCHZ_b_qC_k_ORdCkiPCpFsw",
-    "baosuckhoedoisongboyte":"UCD09NCZlzMj4yGvGkLbuBpw",
-}
-def _yt_rss_shorts(channel, count=24):
-    """LIVE newest uploads via the YouTube RSS feed (stable from datacenter IPs, no API key).
-    Returns newest-first items. This is the primary, reliable source for the shorts slider."""
-    cid=_YT_CHANNEL_IDS.get(channel)
-    if not cid:return[]
-    try:
-        url=f"https://www.youtube.com/feeds/videos.xml?channel_id={cid}"
-        r=requests.get(url,headers={**HEADERS,"Accept-Language":"vi,en;q=0.8"},
-                       cookies={"CONSENT":"YES+1"},timeout=15)
-        if r.status_code!=200 or not r.text:return[]
-        items=[]
-        # RSS is ordered newest-first; parse each <entry>.
-        for ent in re.findall(r"<entry>(.*?)</entry>",r.text,re.S):
-            mv=re.search(r"<yt:videoId>([A-Za-z0-9_-]{11})</yt:videoId>",ent)
-            if not mv:continue
-            vid=mv.group(1)
-            mt=re.search(r"<title>(.*?)</title>",ent,re.S)
-            title=html_lib.unescape(re.sub(r"\s+"," ",mt.group(1)).strip()) if mt else "YouTube Short"
-            items.append({"title":title,"link":f"https://www.youtube.com/watch?v={vid}",
-                          "img":f"https://i.ytimg.com/vi/{vid}/hqdefault.jpg","source":"yt","id":vid,"channel":channel})
-            if len(items)>=count:break
-        return items
-    except:return[]
-def _yt_jina_page(handle, path):
-    """Fetch a YouTube channel tab (videos/shorts) as markdown via r.jina.ai.
-    YouTube serves an EMPTY feeds/videos.xml to datacenter IPs, but the rendered
-    channel page through Jina returns the real newest videos (with Vietnamese titles
-    when X-Locale=vi-VN). This is the reliable source from the HF Space."""
-    u=f"https://www.youtube.com/@{handle}/{path}"
-    try:
-        r=requests.get("https://r.jina.ai/"+u,
-                       headers={"User-Agent":"Mozilla/5.0","X-Return-Format":"markdown","X-Locale":"vi-VN"},
-                       timeout=40)
-        if r.status_code!=200:return ""
-        return r.text
-    except:return ""
-def _yt_jina_shorts(channel, count=24):
-    """LIVE newest videos for a channel via Jina-rendered channel pages (works from the Space).
-    Pulls /videos then /shorts, newest-first, original Vietnamese titles."""
-    items=[];seen=set()
-    for path in ("videos","shorts"):
-        md=_yt_jina_page(channel,path)
-        if not md:continue
-        for m in re.finditer(r'\[([^\]\[]{8,160})\]\(https?://(?:www\.)?youtube\.com/(?:shorts/|watch\?v=)([A-Za-z0-9_-]{11})[^)]*\)',md):
-            title=html_lib.unescape(m.group(1)).strip().strip('"')
-            vid=m.group(2)
-            if vid in seen:continue
-            tl=title.lower()
-            if tl in ("home home","shorts shorts","image") or title.startswith("!") or title.startswith("Image"):continue
-            seen.add(vid)
-            items.append({"title":title,"link":f"https://www.youtube.com/watch?v={vid}",
-                          "img":f"https://i.ytimg.com/vi/{vid}/hqdefault.jpg","source":"yt","id":vid,"channel":channel})
-            if len(items)>=count:break
-        if len(items)>=count:break
-    return items
-def _yt_channel_shorts(channel, count=24):
-    """Live newest videos. Prefer Jina-rendered channel pages (reliable from datacenter IPs),
-    then the RSS feed, then a raw /shorts HTML scrape."""
-    items=_yt_jina_shorts(channel,count)
-    if items:return items
-    items=_yt_rss_shorts(channel,count)
-    if items:return items
-    # Fallback: HTML scrape of the shorts tab.
-    try:
-        url=f"https://www.youtube.com/@{channel}/shorts"
-        r=requests.get(url,headers={**HEADERS,"Accept-Language":"vi,en;q=0.8"},
-                       cookies={"CONSENT":"YES+1"},timeout=15)
-        if r.status_code!=200:return[]
-        html=r.text
-        ids=[];items=[]
-        for m in re.finditer(r'"videoId":"([A-Za-z0-9_-]{11})"',html):
-            vid=m.group(1)
-            if vid in ids:continue
-            ids.append(vid)
-            snip=html[max(0,m.start()-900):m.start()+1600]
-            title=""
-            mt=re.search(r'"title":\{"runs":\[\{"text":"([^"]+)"',snip)
-            if not mt:mt=re.search(r'"accessibilityText":"([^"]+)"',snip)
-            if mt:title=html_lib.unescape(mt.group(1)).replace('\n',' ').strip()
-            if not title:title="YouTube Short"
-            items.append({"title":title,"link":f"https://www.youtube.com/watch?v={vid}","img":f"https://i.ytimg.com/vi/{vid}/hqdefault.jpg","source":"yt","id":vid,"channel":channel})
-            if len(items)>=count:break
-        return items
-    except:return[]
-# ----- Short vs normal-video classification -----
-# A real YouTube Short responds 200 at /shorts/<id>; a normal (horizontal) video
-# 3xx-redirects to /watch?v=<id>. We cache the result per id (it never changes).
-_SHORT_KIND_CACHE={}
-_SHORT_KIND_LOCK=threading.Lock()
-def _classify_short(vid):
-    """Return True if <vid> is a real vertical Short, False if it's a normal video.
-    Cached forever per id; network failures default to True (keep it in the shorts feed)."""
-    if not vid:return True
-    with _SHORT_KIND_LOCK:
-        if vid in _SHORT_KIND_CACHE:return _SHORT_KIND_CACHE[vid]
-    is_short=True
-    try:
-        r=requests.head(f"https://www.youtube.com/shorts/{vid}",
-                        headers={**HEADERS,"Accept-Language":"vi,en;q=0.8"},
-                        cookies={"CONSENT":"YES+1"},allow_redirects=False,timeout=4)
-        # 200 -> short ; 30x to /watch -> normal video
-        if r.status_code in (301,302,303,307,308):
-            loc=r.headers.get("Location","")
-            is_short = "/shorts/" in loc   # redirect that stays in /shorts is still a short
-            if "/watch" in loc:is_short=False
-        elif r.status_code==200:
-            is_short=True
-    except:pass
-    with _SHORT_KIND_LOCK:_SHORT_KIND_CACHE[vid]=is_short
-    return is_short
-def _classify_many(vids):
-    """Classify a list of video ids in parallel; updates each dict's is_short in place."""
-    todo=[v for v in vids if v.get("id") and "is_short" not in v]
-    if not todo:return
-    with ThreadPoolExecutor(min(8,len(todo))) as ex:
-        futs={ex.submit(_classify_short,v["id"]):v for v in todo}
-        for f in as_completed(futs):
-            v=futs[f]
-            try:v["is_short"]=f.result()
-            except:v["is_short"]=True
-def scrape_shorts():
-    """Stable feed of Dân trí & SKĐS YouTube videos, each tagged is_short (true=short, false=normal).
-    Fast RSS scrape + static fallback so the slider never disappears."""
-    vids=[]
-    with ThreadPoolExecutor(2) as ex:
-        futs=[ex.submit(_yt_channel_shorts,ch,24) for ch in ["baodantri7941","baosuckhoedoisongboyte"]]
-        for f in as_completed(futs):
-            try:
-                r=f.result()
-                if r:vids.extend(r)
-            except:pass
-    merged=[];seen=set()
-    for v in vids+SHORTS_FALLBACK:
-        vid=v.get("id")
-        if not vid or vid in seen:continue
-        seen.add(vid);merged.append(v)
-    merged=merged[:40]
-    _classify_many(merged)  # tag each with is_short
-    return merged
-
-# ===== LIVESCORE =====
-@app.get("/api/livescore/live")
-def api_livescore_live():return JSONResponse({"html":_cached("ls_live",lambda:fetch_bongda_api("/api/fixtures/live"),ttl=_cache_ttl_live)})
-@app.get("/api/livescore/incoming")
-def api_livescore_incoming():return JSONResponse({"html":_cached("ls_incoming",lambda:fetch_bongda_api("/api/fixtures/incoming"),ttl=_cache_ttl_live)})
-@app.get("/api/livescore/today")
-def api_livescore_today():
-    today=datetime.now().strftime("%Y-%m-%d");return JSONResponse({"html":_cached("ls_today",lambda:fetch_bongda_api(f"/api/fixtures/get-by-date?date={today}"),ttl=_cache_ttl)})
-@app.get("/api/livescore/results")
-def api_livescore_results():
-    today=datetime.now().strftime("%Y-%m-%d");return JSONResponse({"html":_cached("ls_results",lambda:fetch_bongda_api(f"/api/fixtures/get-by-date?date={today}&status=finished"),ttl=_cache_ttl)})
-@app.get("/api/livescore/standings/{league}")
-def api_livescore_standings(league:str):
-    tid=LEAGUE_IDS.get(league,27110);return JSONResponse({"html":_cached(f"ls_bxh_{league}",lambda:fetch_bongda_api(f"/api/league-table/home?tournament_id={tid}&is_detail=True"),ttl=_cache_ttl)})
-@app.get("/api/livescore/date/{date}")
-def api_livescore_date(date:str):return JSONResponse({"html":fetch_bongda_api(f"/api/fixtures/get-by-date?date={date}")})
-@app.get("/api/match/{event_id}/commentaries")
-def api_match_commentaries(event_id:int):return JSONResponse({"html":fetch_bongda_api(f"/api/fixtures/commentaries?event_id={event_id}")})
-@app.get("/api/match/{event_id}/stats")
-def api_match_stats(event_id:int):return JSONResponse({"html":fetch_bongda_api(f"/api/event-standing/player-performance?event_id={event_id}")})
-@app.get("/api/livescore/featured")
-def api_livescore_featured():
-    def _f():
-        sources=[("/api/fixtures/live","live"),("/api/fixtures/get-by-date?date="+datetime.now().strftime("%Y-%m-%d"),"today"),("/api/fixtures/incoming","upcoming")]
-        for endpoint, stype in sources:
-            html=fetch_bongda_api(endpoint)
-            if not html or len(html)<100:continue
-            soup=BeautifulSoup(html,"lxml");all_matches=[]
-            for li in soup.select("li.match-detail"):
-                match=_parse_match_from_li(li, stype)
-                if not match or not match["event_id"]:continue
-                if stype=="today" and "KT" in match.get("minute",""):continue
-                all_matches.append(match)
-            if not all_matches:continue
-            for pl in PRIORITY_LEAGUES:
-                for match in all_matches:
-                    if pl in match["league"]:return match
-            return all_matches[0]
-        return None
-    return JSONResponse(_cached("ls_featured",_f,ttl=30))
-
-# ===== VIDEO APIs =====
-@app.get("/api/shorts")
-def api_shorts():return JSONResponse(_cached("yt_shorts_v3",scrape_shorts,ttl=_cache_ttl_yt))
-@app.get("/api/short-stats")
-def api_short_stats(ids:str=Query(default="")):
-    arr=[x for x in ids.split(",") if x]
-    with _short_lock:
-        db=_load_short_db();out={}
-        for vid in arr:
-            st=db.get(vid) or _short_default()
-            out[vid]={"views":int(st.get("views",0)),"likes":int(st.get("likes",0)),"shares":int(st.get("shares",0)),"comments":st.get("comments",[])[:80]}
-    return JSONResponse({"stats":out})
-
-@app.post("/api/short-action")
-async def api_short_action(request:Request):
-    try:body=await request.json()
-    except:body={}
-    vid=str(body.get("id","")).strip();action=str(body.get("action","")).strip();txt=str(body.get("text","")).strip()
-    if not vid:return JSONResponse({"error":"missing id"},status_code=400)
-    with _short_lock:
-        db=_load_short_db();st=db.get(vid) or _short_default()
-        if action=="view":st["views"]=int(st.get("views",0))+1
-        elif action=="like":st["likes"]=int(st.get("likes",0))+1
-        elif action=="share":st["shares"]=int(st.get("shares",0))+1
-        elif action=="comment" and txt:
-            comments=st.get("comments",[])
-            comments.insert(0,{"text":txt[:180],"ts":int(time.time())})
-            st["comments"]=comments[:80]
-        st["updated"]=int(time.time());db[vid]=st;_save_short_db(db)
-        out={"views":int(st.get("views",0)),"likes":int(st.get("likes",0)),"shares":int(st.get("shares",0)),"comments":st.get("comments",[])[:80]}
-    return JSONResponse({"stats":out})
-
-@app.get("/api/highlights")
-def api_highlights():return JSONResponse(_cached("xemlaibongda_hl",scrape_xemlaibongda,ttl=_cache_ttl))
-@app.get("/api/highlights/leagues")
-def api_highlights_leagues():return JSONResponse(_cached("hl_leagues",scrape_all_league_highlights,ttl=_cache_ttl))
-@app.get("/api/highlights/{league}")
-def api_highlights_league(league:str):
-    if league not in HL_LEAGUES:return JSONResponse({"error":"league not found"})
-    return JSONResponse(_cached(f"hl_{league}",lambda:scrape_highlights_by_league(league),ttl=_cache_ttl))
-@app.get("/api/highlights_config")
-def api_highlights_config():return JSONResponse(HL_LEAGUES)
-@app.get("/api/video_url")
-def api_video_url(url:str=Query(...)):
-    if "youtube.com" in url or "youtu.be" in url:
-        m=re.search(r'(?:v=|shorts/|youtu\.be/)([a-zA-Z0-9_-]{11})',url)
-        if m:vid=m.group(1);return JSONResponse({"src":f"https://www.youtube.com/embed/{vid}?autoplay=1&rel=0&enablejsapi=1","poster":f"https://i.ytimg.com/vi/{vid}/hqdefault.jpg","type":"youtube"})
-    if "xemlaibongda.top" in url:
-        v=extract_xemlaibongda_video(url)
-        if v:
-            if v["type"]=="hls":v["src"]="/api/proxy/m3u8?url="+quote(v["src"],safe="")
-            return JSONResponse(v)
-    if "bongdaplus.vn" in url:
-        try:
-            m=re.search(r'-(\d{6,})\.html',url)
-            if m:
-                r=requests.get(f"{BASE_BDP}/video-embed/{m.group(1)}.html",headers=HEADERS,timeout=10);r.encoding="utf-8"
-                soup=BeautifulSoup(r.text,"lxml");video=soup.select_one("video#videoPlayer")
-                if video:
-                    source=video.find("source");src=source.get("src","") if source else "";poster=video.get("poster","")
-                    if src:return JSONResponse({"src":"/api/proxy/video?url="+quote(src,safe=""),"poster":poster,"type":"video"})
-        except:pass
-    return JSONResponse({"error":"not found"})
-@app.get("/api/bdp_videos")
-def api_bdp_videos():
-    def _f():
-        try:
-            soup=_get(f"{BASE_BDP}/video");arts=[];seen=set()
-            for a in soup.find_all("a",href=True):
-                href=a.get("href","")
-                if"/video/" not in href or href in("/video/","/video/ban-thang-dep","/video/highlight"):continue
-                if not href.startswith("http"):href=BASE_BDP+href
-                if href in seen:continue
-                title=re.sub(r'^\d{2}:\d{2}','',a.get_text(strip=True)).strip()
-                if not title or len(title)<5:continue
-                img_tag=a.find("img") or(a.parent.find("img") if a.parent else None)
-                img=(img_tag.get("data-src") or img_tag.get("src","")) if img_tag else ""
-                seen.add(href);arts.append({"title":title,"link":href,"img":img,"source":"bdp"})
-            return arts[:20]
-        except:return[]
-    return JSONResponse(_cached("bdp_videos",_f))
-# ===== NEWS =====
-def scrape_vne(cat_url):
-    try:
-        soup=_get(cat_url);arts=[]
-        for it in soup.select("article.item-news")[:15]:
-            a=it.select_one("h2.title-news a") or it.select_one("h3.title-news a")
-            if not a:continue
-            t=a.get("title","") or a.get_text(strip=True);lk=a.get("href","")
-            if not t or not lk:continue
-            im=it.find("img");img=(im.get("data-src") or im.get("src","")) if im else ""
-            if img and'blank'in img:
-                src=it.find("source")
-                if src:img=src.get("srcset","").split(",")[0].strip().split(" ")[0]
-            arts.append({"title":t,"link":lk,"img":img,"source":"vne"})
-        return arts
-    except:return[]
-def scrape_vne_article(url):
-    try:
-        soup=_get(url);h1=soup.select_one("h1.title-detail");desc=soup.select_one("p.description")
-        og=soup.find("meta",property="og:image");og_img=og.get("content","") if og else ""
-        cd=soup.select_one("article.fck_detail");body=[]
-        if cd:
-            for ch in cd.children:
-                if not hasattr(ch,'name') or not ch.name:continue
-                if ch.name=="p":t=ch.get_text(strip=True);(body.append({"type":"p","text":t}) if t else None)
-                elif ch.name=="figure":
-                    im=ch.find("img")
-                    if im:s=im.get("data-src") or im.get("src","");body.append({"type":"img","src":s})
-                elif ch.name in("h2","h3"):body.append({"type":"heading","text":ch.get_text(strip=True)})
-        return{"title":h1.get_text(strip=True) if h1 else "","summary":desc.get_text(strip=True) if desc else "","og_image":og_img,"body":body,"source":"vne","url":url}
-    except:return None
-def _scrape_dantri_homepage(cat_filter=None):
-    try:
-        soup=_get("https://dantri.com.vn/");arts=[];seen=set()
-        for a in soup.find_all("a",href=True):
-            href=a.get("href","");title=a.get("title","") or a.get_text(strip=True)
-            if not title or len(title)<15 or"javascript:" in href:continue
-            if not href.startswith("http"):href="https://dantri.com.vn"+href
-            if href in seen or not href.endswith(".htm"):continue
-            if cat_filter and f"/{cat_filter}/" not in href:continue
-            img_tag=a.find("img")
-            if not img_tag and a.parent:img_tag=a.parent.find("img")
-            img_src=""
-            if img_tag:img_src=img_tag.get("data-src","") or img_tag.get("src","")
-            if not img_src or "cdn" not in img_src:continue
-            proxied_img="/api/proxy/img?url="+quote(img_src,safe="")
-            seen.add(href);arts.append({"title":title,"link":href,"img":proxied_img,"source":"dantri"})
-            if len(arts)>=15:break
-        return arts
-    except:return[]
-def scrape_dantri_hot():return _scrape_dantri_homepage()
-def scrape_dantri_congnghe():
-    try:
-        soup=_get("https://dantri.com.vn/");arts=[];seen=set()
-        for a in soup.find_all("a",href=True):
-            href=a.get("href","");title=a.get("title","") or a.get_text(strip=True)
-            if not title or len(title)<15 or"javascript:" in href:continue
-            if not href.startswith("http"):href="https://dantri.com.vn"+href
-            if href in seen or not href.endswith(".htm"):continue
-            if"/cong-nghe/" not in href:continue
-            img_tag=a.find("img")
-            if not img_tag and a.parent:img_tag=a.parent.find("img")
-            img_src=""
-            if img_tag:img_src=img_tag.get("data-src","") or img_tag.get("src","")
-            if img_src and "cdn" in img_src:img_src="/api/proxy/img?url="+quote(img_src,safe="")
-            else:img_src=""
-            seen.add(href);arts.append({"title":title,"link":href,"img":img_src,"source":"dantri"})
-            if len(arts)>=15:break
-        return arts
-    except:return[]
-
-# ===== JINA READER (for Cloudflare/SPA sites: vatvostudio, bongda, bongdaplus) =====
-_JINA_CACHE={}
-_JINA_LOCK=threading.Lock()
-def _jina_md(url, ttl=600):
-    """Fetch a page as markdown via r.jina.ai (bypasses Cloudflare/JS SPA). Cached."""
-    now=time.time()
-    with _JINA_LOCK:
-        c=_JINA_CACHE.get(url)
-        if c and now-c[0]<ttl:return c[1]
-    md=""
-    try:
-        r=requests.get("https://r.jina.ai/"+url,
-            headers={"Accept":"text/markdown","X-Return-Format":"markdown","User-Agent":"Mozilla/5.0"},timeout=30)
-        if r.status_code==200:md=r.text
-    except:pass
-    with _JINA_LOCK:_JINA_CACHE[url]=(now,md)
-    return md
-
-def _jina_articles(md, host_re, require_id=True):
-    """Parse markdown -> [{title,url,img}] for a given host regex.
-    Handles two markdown forms Jina emits:
-      [![Image..](IMG)](URL)   -> gives image + url (title often empty here)
-      [Plain Title](URL)       -> gives title + url
-    Merge both by URL so each article has a title AND (when available) an image."""
-    titles={}; imgs={}; order=[]
-    def _ok(u):
-        u=u.split(' ')[0].strip()
-        if u.endswith(('.png','.jpg','.jpeg','.webp','.svg','.gif')):return None
-        if require_id and not re.search(r'-[a-z]?\d{4,}\.html', u):return None
-        if any(x in u for x in('/category/','/author/','/tag/','/page/')):return None
-        if not re.match('^'+host_re, u):return None
-        return u
-    # 1) image-wrapped links: [![Image...](IMG)](URL "optional title")
-    for m in re.finditer(r'\[!\[[^\]]*\]\((https?://[^)\s]+)\)\]\(('+host_re+r'[^)\s]+)(?:\s+"[^"]*")?\)', md):
-        u=_ok(m.group(2))
-        if not u:continue
-        if u not in imgs:imgs[u]=m.group(1)
-        if u not in order:order.append(u)
-    # 2) plain text links: [Title](URL "optional title")
-    for m in re.finditer(r'(?<!!)\[([^\]\[]{15,140})\]\(('+host_re+r'[^)\s]+)(?:\s+"[^"]*")?\)', md):
-        title=m.group(1).strip().strip('"')
-        if title.startswith('!') or title.lower() in ('image','tin tức','trang chủ'):continue
-        u=_ok(m.group(2))
-        if not u:continue
-        if u not in titles:titles[u]=title
-        if u not in order:order.append(u)
-    out=[]
-    for u in order:
-        title=titles.get(u,'')
-        if not title or len(title)<15:continue
-        out.append({"title":title,"link":u,"img":imgs.get(u,"")})
-    return out
-
-def scrape_vatvostudio():
-    """Vat Vo Studio (vatvostudio.vn) - Cloudflare-protected WP tech/AI site via Jina. Newest first."""
-    md=_jina_md("https://vatvostudio.vn/")
-    arts=[]
-    for a in _jina_articles(md, r'https://vatvostudio\.vn/', require_id=False):
-        u=a["link"]
-        # real articles are /<slug>/ (one path segment, ends with /)
-        if not re.match(r'^https://vatvostudio\.vn/[a-z0-9\-]{8,}/?$', u):continue
-        a["source"]="vatvo";arts.append(a)
-        if len(arts)>=24:break
-    return arts
-
-def scrape_bongda_jina(limit=30):
-    """bongda.com.vn homepage articles (via Jina; site is JS-protected)."""
-    md=_jina_md("https://bongda.com.vn/")
-    arts=[a for a in _jina_articles(md, r'https://bongda\.com\.vn/')]
-    for a in arts:a["source"]="bongda";a.setdefault("img","")
-    return arts[:limit]
-
-def scrape_bongdaplus_jina(limit=30):
-    """bongdaplus.vn homepage articles (via Jina; site is a JS SPA)."""
-    md=_jina_md("https://bongdaplus.vn/")
-    arts=[a for a in _jina_articles(md, r'https://bongdaplus\.vn/')]
-    for a in arts:a["source"]="bongdaplus";a.setdefault("img","")
-    return arts[:limit]
-
-def scrape_genk_ai():
-    """Scrape AI articles from genk.vn - readable in-app"""
-    try:
-        r=requests.get("https://genk.vn/ai.chn",headers=HEADERS,timeout=15)
-        if r.status_code!=200:return[]
-        r.encoding="utf-8";soup=BeautifulSoup(r.text,"lxml")
-        articles=[];seen=set()
-        for a in soup.find_all("a",href=True):
-            href=a.get("href","")
-            if not href.endswith(".chn") or href=="/ai.chn":continue
-            if href.startswith("/"):href="https://genk.vn"+href
-            if href in seen or "genk.vn" not in href:continue
-            title=a.get("title","") or a.get_text(strip=True)
-            if not title or len(title)<20:continue
-            container=a.parent;img_src=""
-            for _ in range(6):
-                if container is None:break
-                for img in container.find_all("img"):
-                    s=img.get("data-src","") or img.get("src","")
-                    if s and "mediacdn" in s and "avatar" not in s and "logo" not in s:
-                        img_src=s;break
-                if img_src:break
-                container=container.parent
-            seen.add(href)
-            if not img_src:
-                try:
-                    og_r=requests.get(href,headers=HEADERS,timeout=8);og_r.encoding="utf-8"
-                    og_soup=BeautifulSoup(og_r.text,"lxml");og_tag=og_soup.find("meta",property="og:image")
-                    if og_tag:img_src=og_tag.get("content","")
-                except:pass
-            articles.append({"title":title,"link":href,"img":img_src,"source":"genk"})
-            if len(articles)>=30:break
-        return articles
-    except:return[]
-
-def scrape_dantri_article(url):
-    try:
-        r=requests.get(url,headers=HEADERS,timeout=15);r.encoding="utf-8";soup=BeautifulSoup(r.text,"lxml")
-        for tag in soup.find_all(["script","style","nav","footer","aside"]):tag.decompose()
-        h1=soup.find("h1");og=soup.find("meta",property="og:image");og_img=og.get("content","") if og else ""
-        if og_img and "cdnphoto.dantri" in og_img:og_img="/api/proxy/img?url="+quote(og_img,safe="")
-        content=soup.select_one("main") or soup.select_one("div.singular-content") or soup.select_one("article");body=[]
-        if content:
-            for el in content.find_all(["p","h2","h3","figure","img"],recursive=True):
-                if el.name=="p":t=el.get_text(strip=True);(body.append({"type":"p","text":t}) if t and len(t)>15 else None)
-                elif el.name in("h2","h3"):t=el.get_text(strip=True);(body.append({"type":"heading","text":t}) if t else None)
-                elif el.name in("figure","img"):
-                    im=el if el.name=="img" else el.find("img")
-                    if im:
-                        s=im.get("data-src") or im.get("src","")
-                        if s and"base64" not in s:
-                            if "cdnphoto.dantri" in s:s="/api/proxy/img?url="+quote(s,safe="")
-                            body.append({"type":"img","src":s})
-        desc="";sapo=soup.select_one("h2.singular-sapo") or soup.select_one("h2[class*=sapo]")
-        if not sapo:
-            og_desc=soup.find("meta",property="og:description")
-            if og_desc:desc=og_desc.get("content","")
-        else:desc=sapo.get_text(strip=True)
-        return{"title":h1.get_text(strip=True) if h1 else "","summary":desc,"og_image":og_img,"body":body,"source":"dantri","url":url}
-    except:return None
-def scrape_bbc_vietnamese():
-    try:
-        r=requests.get("https://www.bbc.com/vietnamese",headers={"User-Agent":"Mozilla/5.0","Accept-Language":"en-GB"},timeout=15);r.encoding="utf-8"
-        soup=BeautifulSoup(r.text,"lxml");arts=[];seen=set()
-        for a in soup.select("a[href*='/vietnamese/']"):
-            href=a.get("href","")
-            if not href or href=="/vietnamese" or href.count("/")<3:continue
-            if not href.startswith("http"):href="https://www.bbc.com"+href
-            if href in seen:continue
-            title=a.get_text(strip=True)
-            if not title or len(title)<15 or any(x in title.lower() for x in["đăng nhập","trang chủ","bbc news"]):continue
-            img="";container=a.parent
-            for _ in range(3):
-                if container:
-                    im=container.find("img")
-                    if im:img=im.get("src","") or im.get("data-src","");break
-                    container=container.parent
-            seen.add(href);arts.append({"title":title,"link":href,"img":img,"source":"bbc"})
-            if len(arts)>=15:break
-        return arts
-    except:return[]
-def scrape_bbc_article(url):
-    try:
-        r=requests.get(url,headers={"User-Agent":"Mozilla/5.0","Accept-Language":"en-GB"},timeout=15);r.encoding="utf-8"
-        soup=BeautifulSoup(r.text,"lxml");h1=soup.find("h1")
-        og=soup.find("meta",property="og:image");og_img=og.get("content","") if og else ""
-        body=[]
-        for p in soup.select("[data-component='text-block'] p, article p, main p"):
-            t=p.get_text(strip=True)
-            if t and len(t)>20:body.append({"type":"p","text":t})
-        return{"title":h1.get_text(strip=True) if h1 else "","summary":"","og_image":og_img,"body":body,"source":"bbc","url":url}
-    except:return None
-
-def scrape_ttvh_worldcup():
-    """Scrape all World Cup 2026 articles from The Thao Van Hoa RSS."""
-    try:
-        r=requests.get("https://thethaovanhoa.vn/rss/world-cup-2026.rss",headers=HEADERS,timeout=15);r.encoding="utf-8"
-        soup=BeautifulSoup(r.text,"xml");arts=[];seen=set()
-        for it in soup.find_all("item"):
-            title=(it.find("title").get_text(strip=True) if it.find("title") else "")
-            link=(it.find("link").get_text(strip=True) if it.find("link") else "")
-            desc=(it.find("description").get_text(" ",strip=True) if it.find("description") else "")
-            img="";ds=BeautifulSoup(desc,"lxml");im=ds.find("img")
-            if im:img=im.get("src","") or im.get("data-src","")
-            if title and link and link not in seen:
-                seen.add(link);arts.append({"title":title,"link":link,"img":img,"source":"ttvh"})
-        if arts:return arts
-    except:pass
-    try:
-        soup=_get("https://thethaovanhoa.vn/world-cup-2026.htm");arts=[];seen=set()
-        for a in soup.find_all("a",href=True):
-            href=a.get("href","")
-            if not href.startswith("http"):href="https://thethaovanhoa.vn"+href
-            if href in seen or "thethaovanhoa.vn" not in href:continue
-            if not re.search(r"/[^/]+-\d{8,}\.htm",href):continue
-            title=a.get("title","") or a.get_text(" ",strip=True)
-            img=None;p=a
-            for _ in range(5):
-                if p is None:break
-                img=p.find("img")
-                if img:break
-                p=p.parent
-            img_src=""
-            if img:
-                img_src=img.get("data-src","") or img.get("src","") or img.get("data-original","") or img.get("data-thumb","")
-                if len(title)<15:title=img.get("alt","") or img.get("title","") or title
-            if not title or len(title)<15:continue
-            seen.add(href);arts.append({"title":title,"link":href,"img":img_src,"source":"ttvh"})
-            if len(arts)>=24:break
-        return arts
-    except:return[]
-
-def scrape_ttvh_article(url):
-    try:
-        soup=_get(url);h1=soup.find("h1");og=soup.find("meta",property="og:image");og_img=og.get("content","") if og else ""
-        og_title=soup.find("meta",property="og:title");fallback_title=og_title.get("content","") if og_title else ""
-        desc_el=soup.find("meta",property="og:description");desc=desc_el.get("content","") if desc_el else ""
-        cd=soup.select_one(".detail-content") or soup.select_one(".content-detail") or soup.select_one("article") or soup.select_one("main")
-        body=[]
-        if cd:
-            for el in cd.find_all(["p","h2","h3","figure","img"],recursive=True):
-                if el.name=="p":
-                    t=el.get_text(strip=True)
-                    if t and len(t)>20 and "Theo dõi" not in t:body.append({"type":"p","text":t})
-                elif el.name in ("h2","h3"):
-                    t=el.get_text(strip=True)
-                    if t:body.append({"type":"heading","text":t})
-                elif el.name in ("figure","img"):
-                    im=el if el.name=="img" else el.find("img")
-                    if im:
-                        src=im.get("data-src") or im.get("src","") or im.get("data-original","")
-                        if src and "base64" not in src:body.append({"type":"img","src":src})
-        if not body and desc:body=[{"type":"p","text":desc}]
-        return {"title":h1.get_text(strip=True) if h1 else fallback_title,"summary":desc,"og_image":og_img,"body":body,"source":"ttvh","url":url}
-    except:return None
-
-VNE_CATS={"thoi-su":("https://vnexpress.net/thoi-su","Thời Sự"),"the-gioi":("https://vnexpress.net/the-gioi","Thế Giới"),"kinh-doanh":("https://vnexpress.net/kinh-doanh","Kinh Doanh"),"the-thao":("https://vnexpress.net/the-thao","Thể Thao"),"giai-tri":("https://vnexpress.net/giai-tri","Giải Trí"),"suc-khoe":("https://vnexpress.net/suc-khoe","Sức Khỏe"),"phap-luat":("https://vnexpress.net/phap-luat","Pháp Luật"),"giao-duc":("https://vnexpress.net/giao-duc","Giáo Dục"),"du-lich":("https://vnexpress.net/du-lich","Du Lịch"),"doi-song":("https://vnexpress.net/doi-song","Đời Sống")}
-@app.get("/api/homepage")
-def api_homepage():
-    def _f():
-        articles=[]
-        with ThreadPoolExecutor(12) as ex:
-            futs={ex.submit(scrape_vne,VNE_CATS[k][0]):VNE_CATS[k][1] for k in["thoi-su","the-gioi","kinh-doanh","the-thao","giai-tri","phap-luat","giao-duc","du-lich","doi-song"]}
-            futs[ex.submit(scrape_bbc_vietnamese)]="BBC"
-            for f in as_completed(futs):
-                try:
-                    for a in f.result():a["group"]=futs[f];articles.append(a)
-                except:pass
-        return articles
-    return JSONResponse(_cached("homepage",_f))
-@app.get("/api/category/{cat_id}")
-def api_category(cat_id:str):
-    def _f():
-        if cat_id=="bbc":return scrape_bbc_vietnamese()
-        if cat_id=="cong-nghe":return scrape_genk_ai()
-        if cat_id in VNE_CATS:arts=scrape_vne(VNE_CATS[cat_id][0]);[a.update({"group":VNE_CATS[cat_id][1]}) for a in arts];return arts
-        return[]
-    return JSONResponse(_cached(f"cat_{cat_id}",_f))
-@app.get("/api/categories")
-def api_categories():
-    cats=[{"id":"bbc","name":"BBC Tiếng Việt","source":"bbc"},{"id":"cong-nghe","name":"Công Nghệ","source":"genk"}]
-    for k,(u,n) in VNE_CATS.items():cats.append({"id":k,"name":n,"source":"vne"})
-    return JSONResponse(cats)
-@app.get("/api/dantri_hot")
-def api_dantri_hot():return JSONResponse(_cached("dantri_hot",scrape_dantri_hot))
-@app.get("/api/genk_ai")
-def api_genk_ai():
-    def _ai_apps():
-        # vatvostudio.vn first (newest), then GenK AI — deduped by title
-        vatvo=_cached("vatvo_ai",scrape_vatvostudio,ttl=_cache_ttl)
-        genk=_cached("genk_ai",scrape_genk_ai,ttl=_cache_ttl)
-        out=[];seen=set()
-        for a in (vatvo or [])+(genk or []):
-            key=(a.get("title","") or "").strip().lower()
-            if not key or key in seen:continue
-            seen.add(key);out.append(a)
-        return out
-    return JSONResponse(_cached("ai_apps_merged",_ai_apps,ttl=_cache_ttl))
-@app.get("/api/worldcup2026")
-def api_worldcup2026():return JSONResponse(_cached("ttvh_worldcup",scrape_ttvh_worldcup,ttl=_cache_ttl))
-def scrape_genk_article(url):
-    try:
-        r=requests.get(url,headers=HEADERS,timeout=15);r.encoding="utf-8";soup=BeautifulSoup(r.text,"lxml")
-        h1=soup.find("h1");og=soup.find("meta",property="og:image");og_img=og.get("content","") if og else ""
-        og_title=soup.find("meta",property="og:title");fallback_title=og_title.get("content","") if og_title else ""
-        desc_el=soup.find("meta",property="og:description");desc=desc_el.get("content","") if desc_el else ""
-        cd=soup.select_one(".knc-content");body=[]
-        if cd:
-            for el in cd.find_all(["p","h2","h3","figure","img"],recursive=True):
-                if el.name=="p":t=el.get_text(strip=True);(body.append({"type":"p","text":t}) if t and len(t)>15 else None)
-                elif el.name in("h2","h3"):t=el.get_text(strip=True);(body.append({"type":"heading","text":t}) if t else None)
-                elif el.name in("figure","img"):
-                    im=el if el.name=="img" else el.find("img")
-                    if im:s=im.get("data-src") or im.get("src","");(body.append({"type":"img","src":s}) if s and"base64" not in s else None)
-        return{"title":h1.get_text(strip=True) if h1 else "","summary":desc,"og_image":og_img,"body":body,"source":"genk","url":url}
-    except:return None
-
-@app.get("/api/article")
-def api_article(url:str=Query(...)):
-    if"vnexpress.net" in url:data=scrape_vne_article(url)
-    elif"bbc.com" in url:data=scrape_bbc_article(url)
-    elif"dantri.com.vn" in url:data=scrape_dantri_article(url)
-    elif"genk.vn" in url:data=scrape_genk_article(url)
-    elif"thethaovanhoa.vn" in url:data=scrape_ttvh_article(url)
-    else:data=None
-    return JSONResponse(data if data else{"error":"not supported"})
-def _web_context(topic):
-    """Collect real web/news context for a topic."""
-    bits=[]
-    try:
-        rss="https://news.google.com/rss/search?q="+quote(topic)+"&hl=vi&gl=VN&ceid=VN:vi"
-        r=requests.get(rss,headers=HEADERS,timeout=12);r.encoding="utf-8"
-        soup=BeautifulSoup(r.text,"xml")
-        for it in soup.find_all("item")[:8]:
-            title=it.find("title").get_text(" ",strip=True) if it.find("title") else ""
-            src=it.find("source").get_text(" ",strip=True) if it.find("source") else ""
-            if title:bits.append((title+(" — "+src if src else ""))[:280])
-    except:pass
-    if bits:return "\n".join(bits)
-    try:
-        r=requests.get("https://html.duckduckgo.com/html/?q="+quote(topic),headers=HEADERS,timeout=12);r.encoding="utf-8"
-        soup=BeautifulSoup(r.text,"lxml")
-        for res in soup.select(".result")[:6]:
-            t=res.select_one(".result__title");sn=res.select_one(".result__snippet")
-            line=((t.get_text(" ",strip=True) if t else "")+" — "+(sn.get_text(" ",strip=True) if sn else "")).strip(" —")
-            if line:bits.append(line[:280])
-    except:pass
-    return "\n".join(bits)
-
-def _jina_read(url):
-    try:
-        ju="https://r.jina.ai/http://"+url
-        r=requests.get(ju,headers=HEADERS,timeout=25);r.encoding="utf-8"
-        if r.status_code!=200 or not r.text:return None
-        lines=[x.rstrip() for x in r.text.splitlines()]
-        title="";img="";body=[];summary=""
-        for ln in lines[:40]:
-            if ln.startswith("Title:"):title=ln.replace("Title:","",1).strip()
-            elif ln.startswith("Image:"):img=ln.replace("Image:","",1).strip()
-            elif ln.startswith("Description:"):summary=ln.replace("Description:","",1).strip()
-        for ln in lines:
-            t=ln.strip()
-            if not t or t.startswith(("Title:","URL Source:","Published Time:","Markdown Content:","Image:","Description:")):continue
-            if len(t)>40:body.append({"type":"p","text":t})
-        if not body and summary:body=[{"type":"p","text":summary}]
-        return {"title":title or url,"summary":summary,"og_image":img,"body":body[:80],"source":"jina","url":url}
-    except:return None
-
-def _scrape_generic_article(url):
-    try:
-        hdr={**HEADERS,"Accept":"text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8"}
-        r=requests.get(url,headers=hdr,timeout=15);r.encoding="utf-8"
-        ct=r.headers.get("content-type","").lower()
-        if r.status_code>=400 or "text/html" not in ct:
-            jr=_jina_read(url)
-            if jr:return jr
-        soup=BeautifulSoup(r.text,"lxml")
-        for tag in soup.find_all(["script","style","nav","footer","aside","form"]):tag.decompose()
-        h1=soup.find("h1")
-        ogt=soup.find("meta",property="og:title");title=h1.get_text(strip=True) if h1 else (ogt.get("content","") if ogt else "")
-        ogd=soup.find("meta",property="og:description");desc=ogd.get("content","") if ogd else ""
-        ogi=soup.find("meta",property="og:image");img=ogi.get("content","") if ogi else ""
-        main=soup.find("article") or soup.find("main") or soup.body
-        body=[]
-        if main:
-            for el in main.find_all(["p","h2","h3","figure","img"],recursive=True):
-                if el.name=="p":
-                    t=el.get_text(" ",strip=True)
-                    if t and len(t)>35:body.append({"type":"p","text":t})
-                elif el.name in ("h2","h3"):
-                    t=el.get_text(" ",strip=True)
-                    if t:body.append({"type":"heading","text":t})
-                elif el.name in ("figure","img"):
-                    im=el if el.name=="img" else el.find("img")
-                    if im:
-                        src=im.get("data-src") or im.get("src","") or im.get("data-original","")
-                        if src and "base64" not in src:body.append({"type":"img","src":src})
-        if not body:
-            jr=_jina_read(url)
-            if jr and jr.get("body"):return jr
-        if not body and desc:body=[{"type":"p","text":desc}]
-        return {"title":title or url,"summary":desc,"og_image":img,"body":body,"source":"generic","url":url}
-    except:
-        return _jina_read(url)
-
-def _article_by_url(url):
-    if "vnexpress.net" in url:return scrape_vne_article(url)
-    if "bbc.com" in url:return scrape_bbc_article(url)
-    if "dantri.com.vn" in url:return scrape_dantri_article(url)
-    if "genk.vn" in url:return scrape_genk_article(url)
-    if "thethaovanhoa.vn" in url:return scrape_ttvh_article(url)
-    return _scrape_generic_article(url)
-
-_QWEN_TEXT_MODELS=[m.strip() for m in os.environ.get(
-    "QWEN_TEXT_MODELS",
-    "Qwen/Qwen2.5-72B-Instruct,meta-llama/Llama-3.3-70B-Instruct,Qwen/Qwen2.5-7B-Instruct,meta-llama/Llama-3.1-8B-Instruct"
-).split(",") if m.strip()]
-_QWEN_WORKING=None  # cache last-working model id
-def _call_qwen(prompt, max_tokens=1800):
-    """Summarize via HF router using FAST TEXT models (no vision needed for text summary).
-    Tries text models first (much faster than the VL model), VL only as last resort.
-    Returns None if no model is available (e.g. token lacks inference permission / credits)."""
-    global _QWEN_WORKING
-    token=os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACEHUB_API_TOKEN") or os.environ.get("VAISTUDIO")
-    if not token:return None
-    headers={"Authorization":"Bearer "+token,"Content-Type":"application/json"}
-    models=list(_QWEN_TEXT_MODELS)+["Qwen/Qwen2.5-VL-7B-Instruct"]
-    if _QWEN_WORKING and _QWEN_WORKING in models:
-        models=[_QWEN_WORKING]+[m for m in models if m!=_QWEN_WORKING]
-    last=None
-    for model in models:
-        try:
-            payload={"model":model,"messages":[{"role":"user","content":prompt}],"max_tokens":max_tokens,"temperature":0.7}
-            r=requests.post("https://router.huggingface.co/v1/chat/completions",headers=headers,json=payload,timeout=60)
-            if r.status_code>=300:
-                last=f"{model}:{r.status_code}";continue
-            j=r.json();out=j.get("choices",[{}])[0].get("message",{}).get("content")
-            if out and out.strip():
-                _QWEN_WORKING=model
-                return out
-        except Exception as e:
-            last=f"{model}:{type(e).__name__}";continue
-    if last:print("[_call_qwen] all models failed:",last)
-    return None
-
-def _collect_article_text(data, limit=28000):
-    title=(data or {}).get("title","");summary=(data or {}).get("summary","")
-    parts=[]
-    if summary:parts.append(summary)
-    for b in (data or {}).get("body",[]):
-        if b.get("type")=="heading":parts.append("## "+b.get("text","") )
-        elif b.get("type")=="p":parts.append(b.get("text","") )
-    text="\n".join([p.strip() for p in parts if p and p.strip()])
-    return title,text[:limit]
-
-def _ai_rewrite_article(data,tone="tu-nhien"):
-    title,text=_collect_article_text(data)
-    prompt=("Bạn là biên tập viên báo điện tử tiếng Việt. Hãy viết lại bài dưới đây bằng ngôn ngữ tự nhiên, mạch lạc, không cắt khúc, không bỏ ý quan trọng. "
-            "Giữ đúng sự thật, không bịa, không thêm thông tin ngoài bài. Văn phong: "+tone+". "
-            "Đầu ra gồm: tiêu đề hấp dẫn, đoạn sapo 2-3 câu, các đoạn nội dung ngắn dễ đọc, và 3 gạch đầu dòng điểm chính.\n\n"
-            "TIÊU ĐỀ GỐC: "+title+"\n\nNỘI DUNG GỐC:\n"+text)
-    out=_call_qwen(prompt,2200)
-    if out and len(out)>300:return out.strip()
-    # Fallback: complete non-truncated rewrite using full collected text chunks
-    paras=[p.strip() for p in text.split("\n") if len(p.strip())>30]
-    body="\n\n".join(paras[:18])
-    bullets="\n".join(["• "+p[:220]+("..." if len(p)>220 else "") for p in paras[:5]])
-    return ("Bản tin AI viết lại: "+title+"\n\n"+
-            (paras[0] if paras else "")+"\n\n"+body+"\n\nĐiểm chính:\n"+bullets).strip()
-
-def _image_for_topic(topic):
-    return "https://image.pollinations.ai/prompt/"+quote("editorial illustration, Vietnamese news, "+topic,safe="")+"?width=1024&height=576&nologo=true"
-
-def _topic_articles(topic,limit=5):
-    items=[];seen=set()
-    try:
-        rss="https://news.google.com/rss/search?q="+quote(topic)+"&hl=vi&gl=VN&ceid=VN:vi"
-        r=requests.get(rss,headers=HEADERS,timeout=12);r.encoding="utf-8"
-        soup=BeautifulSoup(r.text,"xml")
-        for it in soup.find_all("item")[:limit*3]:
-            title=it.find("title").get_text(" ",strip=True) if it.find("title") else ""
-            link=it.find("link").get_text(strip=True) if it.find("link") else ""
-            src=it.find("source").get_text(" ",strip=True) if it.find("source") else ""
-            if not title or not link or link in seen:continue
-            seen.add(link);items.append({"title":title,"link":link,"source":src})
-            if len(items)>=limit:break
-    except:pass
-    return items
-
-def _topic_article_context(topic):
-    """Filter readable article sources by topic, then summarize actual article bodies."""
-    raw_keys=[k.lower() for k in re.findall(r"[\wÀ-ỹ]+",topic) if len(k)>2]
-    # Drop ultra-generic tokens; keep domain words such as giáo/dục, bóng/đá, world/cup.
-    stop={"trong","năm","the","and","của","cho","với","một","các","những","hiện","nay"}
-    keys=[k for k in raw_keys if k not in stop]
-    candidates=[];seen=set()
-    def add_items(items):
-        for a in items or []:
-            link=a.get("link","");title=a.get("title","")
-            if not link or link in seen:continue
-            seen.add(link);candidates.append(a)
-    try:add_items(scrape_genk_ai())
-    except:pass
-    try:add_items(scrape_dantri_congnghe())
-    except:pass
-    try:add_items(scrape_ttvh_worldcup())
-    except:pass
-    # PRE-FILTER by TITLE first so we don't waste time scraping 40 irrelevant articles.
-    # Only candidates whose title already matches a topic keyword get fully scraped.
-    if keys:
-        pref=[a for a in candidates if sum(1 for k in keys if k in (a.get("title","").lower()))>0]
-        candidates=pref or candidates
-    candidates=candidates[:12]  # scrape at most 12 (was looping over 40 serially)
-    # Scrape the shortlisted articles CONCURRENTLY (this is the main speed-up for
-    # 'rewrite tong hop nguon tin tu hashtag'). Serial scraping of many articles via
-    # Jina/HTTP was the real cause of the long wait, NOT the LLM call.
-    from concurrent.futures import ThreadPoolExecutor as _TPE
-    def _fetch(a):
-        try:return a,_article_by_url(a.get("link",""))
-        except Exception:return a,None
-    fetched=[]
-    with _TPE(max_workers=8) as ex:
-        for a,data in ex.map(_fetch,candidates):
-            fetched.append((a,data))
-    scored=[];img=""
-    for a,data in fetched:
-        if not data or not data.get("body"):continue
-        title=data.get("title") or a.get("title","")
-        ps=[b.get("text","") for b in data.get("body",[]) if b.get("type")=="p" and len(b.get("text",""))>40]
-        excerpt=" ".join(ps)[:1800] or data.get("summary","")
-        hay=(title+" "+excerpt).lower()
-        score=sum(1 for k in keys if k in hay)
-        # Require topic relevance when we have meaningful keys.
-        if keys and score==0:continue
-        if len(keys)>=2 and score<2 and not any(" ".join(keys[i:i+2]) in hay for i in range(len(keys)-1)):continue
-        scored.append((score,title,a.get("link",""),excerpt,data.get("og_image") or a.get("img","") or ""))
-    scored=sorted(scored,key=lambda x:x[0],reverse=True)[:5]
-    chunks=[];items=[]
-    for score,title,link,excerpt,im in scored:
-        if not img and im:img=im
-        chunks.append("BÀI: "+title+"\nURL: "+link+"\nNỘI DUNG LỌC: "+excerpt)
-        # Structured item per news: title + its own image + a SHORT clean description.
-        # Take the first 1-2 real sentences of the article body (no URLs, no prompt text).
-        desc=_short_desc(excerpt,260)
-        ctitle=html_lib.unescape((title or "").replace("\n"," ")).strip()
-        items.append({"title":ctitle,"img":im or "","desc":desc,"link":link})
-    if chunks:return "\n\n".join(chunks),img,items
-    return _web_context(topic),"",[]
-
-def _short_desc(text,limit=260):
-    """First 1-2 clean sentences from an article excerpt (strip URLs / markdown / boilerplate)."""
-    t=re.sub(r"https?://\S+","",text or "")
-    t=re.sub(r"\[([^\]]*)\]\([^)]*\)",r"\1",t)
-    t=re.sub(r"[#>*`_]+"," ",t)
-    t=re.sub(r"\s+"," ",t).strip()
-    if not t:return ""
-    # cut at sentence boundary near the limit
-    if len(t)<=limit:return t
-    cut=t[:limit]
-    p=max(cut.rfind(". "),cut.rfind("! "),cut.rfind("? "))
-    return (cut[:p+1] if p>80 else cut).strip()
-
-def _strip_prompt_leak(out,topic):
-    """Remove any leaked instruction/prompt phrases from the model output."""
-    if not out:return out
-    bad=["bạn là cây bút","bạn là biên tập","nhiệm vụ","vai trò","đầu ra gồm","theo yêu cầu",
-         "dưới đây là","tôi sẽ","đây là bài","nguồn thực tế","không viết chung chung","chỉ dùng dữ kiện"]
-    lines=[]
-    for ln in out.splitlines():
-        low=ln.strip().lower()
-        if any(b in low for b in bad):continue
-        lines.append(ln)
-    return "\n".join(lines).strip()
-
-def _topic_post_text(topic,ctx=None,items=None):
-    if ctx is None:
-        ctx,_,items=_topic_article_context(topic)
-    prompt=("Bạn là cây bút báo điện tử tiếng Việt. Hãy lọc các thông tin thực tế trong những nguồn dưới đây để viết một bài tóm tắt theo chủ đề: "+topic+
-            ". Không viết chung chung. Chỉ dùng dữ kiện có trong nguồn; nếu nguồn khác nhau thì tổng hợp khách quan. "
-            "Viết liền mạch thành đoạn văn, KHÔNG chèn URL, KHÔNG nhắc lại yêu cầu này.\n\nNGUỒN THỰC TẾ:\n"+ctx)
-    out=_call_qwen(prompt,1800)
-    out=_strip_prompt_leak(out,topic)
-    if out and len(out)>300:return out.strip()
-    # Fallback WITHOUT prompt/URL leakage: build a readable digest from the per-news items.
-    if items:
-        parts=["Tổng hợp tin nổi bật về \""+topic+"\":"]
-        for it in items:
-            t=it.get("title","").strip();d=it.get("desc","").strip()
-            if not t:continue
-            parts.append("• "+t+((" — "+d) if d else ""))
-        return "\n\n".join(parts)
-    if ctx:
-        return "Tổng hợp tin nổi bật về \""+topic+"\". Các nguồn cho thấy chủ đề đang có nhiều diễn biến đáng chú ý."
-    return "Chưa thu thập được dữ liệu đủ rõ cho chủ đề: "+topic
-
-@app.get("/api/wall")
-def api_wall():return JSONResponse({"posts":_load_wall()[:50]})
-
-@app.post("/api/rewrite_share")
-async def api_rewrite_share(request:Request):
-    try:body=await request.json()
-    except:body={}
-    url=str(body.get("url","")).strip();tone=str(body.get("tone","tu-nhien")).strip()
-    if not url:return JSONResponse({"error":"missing url"},status_code=400)
-    data=_article_by_url(url)
-    if not data or not data.get("title") or (not data.get("body") and not data.get("summary")):
-        return JSONResponse({"error":"Không đọc được bài viết"},status_code=422)
-    post={"id":hashlib.md5((url+str(time.time())).encode()).hexdigest()[:12],"url":url,"title":data.get("title",""),"img":data.get("og_image","") or "","text":_ai_rewrite_article(data,tone),"ts":int(time.time()),"source":data.get("source","")}
-    posts=_load_wall();posts.insert(0,post);_save_wall(posts)
-    return JSONResponse({"post":post})
-
-@app.post("/api/topic_post")
-async def api_topic_post(request:Request):
-    try:body=await request.json()
-    except:body={}
-    topic=str(body.get("topic","")).strip()
-    if not topic:return JSONResponse({"error":"missing topic"},status_code=400)
-    # Compute the aggregated context ONCE (was computed twice = double scraping work).
-    ctx,ctx_img,items=_topic_article_context(topic)
-    post={"id":hashlib.md5((topic+str(time.time())).encode()).hexdigest()[:12],"url":"","title":topic,
-          "img":ctx_img or (items[0]["img"] if items and items[0].get("img") else "") or _image_for_topic(topic),
-          "text":_topic_post_text(topic,ctx,items),
-          "items":items,  # per-news: each {title, img, desc} (no URLs shown)
-          "ts":int(time.time()),"source":"ai-topic"}
-    posts=_load_wall();posts.insert(0,post);_save_wall(posts)
-    return JSONResponse({"post":post})
-
-@app.post("/api/url_wall")
-async def api_url_wall(request:Request):
-    try:body=await request.json()
-    except:body={}
-    url=str(body.get("url","")).strip()
-    if not url:return JSONResponse({"error":"missing url"},status_code=400)
-    data=_article_by_url(url)
-    if not data or not data.get("title"):
-        return JSONResponse({"error":"Không đọc được URL"},status_code=422)
-    post={"id":hashlib.md5((url+str(time.time())).encode()).hexdigest()[:12],"url":url,"title":data.get("title",""),"img":data.get("og_image","") or "","text":_ai_rewrite_article(data,"ngan-gon-tu-nhien"),"ts":int(time.time()),"source":data.get("source","")}
-    posts=_load_wall();posts.insert(0,post);_save_wall(posts)
-    return JSONResponse({"post":post})
-
-def _share_html(title,img,desc,store_key,store_val):
-    """A share landing page with REAL Open Graph tags (title/image/description) for
-    social previews, that then opens the correct content inside the SPA."""
-    import html as _h,json as _json
-    t=_h.escape((title or "VNEWS").strip());d=_h.escape((desc or "Tin tức tổng hợp, bóng đá, video, AI.").strip()[:200])
-    im=(img or "").strip() or "https://s1.vnecdn.net/vnexpress/restruct/i/v9505/logo_default.jpg"
-    im_e=_h.escape(im,quote=True)
-    val=_json.dumps(store_val)  # safe JS string/obj literal
-    script=(f'<script>try{{localStorage.setItem("{store_key}",{val});}}catch(e){{}}'
-            f'location.replace("{SPACE_URL}/");</script>') if store_val else f'<script>location.replace("{SPACE_URL}/");</script>'
-    return HTMLResponse(
-        '<!DOCTYPE html><html lang="vi"><head><meta charset="utf-8">'
-        '<meta name="viewport" content="width=device-width,initial-scale=1">'
-        f'<title>{t}</title>'
-        f'<meta name="description" content="{d}">'
-        '<meta property="og:type" content="article">'
-        f'<meta property="og:title" content="{t}">'
-        f'<meta property="og:description" content="{d}">'
-        f'<meta property="og:image" content="{im_e}">'
-        '<meta name="twitter:card" content="summary_large_image">'
-        f'<meta name="twitter:title" content="{t}">'
-        f'<meta name="twitter:description" content="{d}">'
-        f'<meta name="twitter:image" content="{im_e}">'
-        f'</head><body style="background:#111;color:#fff;text-align:center;padding:40px;font-family:sans-serif">'
-        f'<img src="{im_e}" style="max-width:320px;border-radius:10px"><h3>{t}</h3><p>{d}</p>'
-        f'<p style="color:#888">⏳ Đang mở…</p>{script}</body></html>')
-
-@app.get("/v")
-async def video_share(url:str=Query(default=""),title:str=Query(default="VNEWS Video"),img:str=Query(default=""),desc:str=Query(default=""),type:str=Query(default="highlights")):
-    decoded_url=unquote(url)
-    val={"url":decoded_url,"type":type} if decoded_url else None
-    return _share_html(unquote(title),unquote(img),unquote(desc),"pending_video",val)
-
-@app.get("/s")
-async def share_redirect(url:str=Query(default=""),title:str=Query(default="VNEWS"),img:str=Query(default=""),desc:str=Query(default=""),post:str=Query(default="")):
-    # Sharing a saved wall/AI post by id, OR an external article url.
-    if post:
-        return _share_html(unquote(title),unquote(img),unquote(desc),"pending_post",unquote(post))
-    decoded_url=unquote(url)
-    return _share_html(unquote(title),unquote(img),unquote(desc),"pending_article",decoded_url if decoded_url else None)
-@app.get("/")
-async def index():
-    with open("/app/static/index.html","r",encoding="utf-8") as f:return HTMLResponse(content=f.read())
-app.mount("/static",StaticFiles(directory="/app/static"),name="static")
+see file /app/main.py