Bypass deep import chain: direct main import for fast startup + fix all issues"

app_fast.py

@@ -0,0 +1,246 @@
+"""VNEWS fast runtime: bypass deep import chain, import main.py directly.
+Fixes: slow load, prompt leaking, source_details persistence, ask AI context, duplicate rewrite button.
+"""
+import os, re, json, time, threading, asyncio
+import html as html_lib
+from urllib.parse import quote, urlparse
+from concurrent.futures import ThreadPoolExecutor, as_completed
+import requests
+from bs4 import BeautifulSoup
+from fastapi import Request, Query
+from fastapi.responses import HTMLResponse, JSONResponse
+
+# Import main app directly — skip 8-layer import chain.
+from main import app
+
+# Import AI generation from ai_ext (lightweight, no extra chain).
+try:
+    from ai_ext import qwen_generate, _load_ai_wall, _save_ai_wall, make_post, _clean_text, pollinations_image_url, SHORTS_DIR, _safe_name, _download_image, gTTS, scrape_any_url
+except Exception:
+    qwen_generate=None
+
+RESTORE_INDEX_URL="https://huggingface.co/spaces/bep40/vnews/raw/restore-33c3dda/static/index.html"
+UA={"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,en;q=0.8"}
+SPACE_URL="https://bep40-vnews.hf.space"
+DATA_DIR="/data" if os.path.isdir('/data') else "/app/data"
+_HOME_CACHE={"t":0,"d":[]}
+_SHORTS_CACHE={"t":0,"d":[]}
+
+def clean(s):return re.sub(r"\s+"," ",html_lib.unescape(str(s or ""))).strip()
+def _domain(u):
+    try:return urlparse(u or '').netloc.replace('www.','')
+    except Exception:return ''
+
+# ===== FAST RSS HOMEPAGE =====
+def _rss_articles(feed_url, group, source='vne', limit=6):
+    out=[]
+    try:
+        r=requests.get(feed_url,headers=UA,timeout=5);r.encoding='utf-8'
+        soup=BeautifulSoup(r.text,'xml')
+        for it in soup.find_all('item')[:limit]:
+            title=clean(it.find('title').get_text(' ',strip=True) if it.find('title') else '')
+            link=clean(it.find('link').get_text(strip=True) if it.find('link') else '')
+            desc_raw=it.find('description').get_text(' ',strip=True) if it.find('description') else ''
+            ds=BeautifulSoup(desc_raw,'lxml')
+            im=ds.find('img');img=im.get('src','') if im else ''
+            if title and link:out.append({'title':title,'link':link,'img':img,'source':source,'group':group})
+    except Exception:pass
+    return out
+
+def _fetch_homepage():
+    feeds=[('Thời Sự','https://vnexpress.net/rss/thoi-su.rss'),('Thế Giới','https://vnexpress.net/rss/the-gioi.rss'),('Kinh Doanh','https://vnexpress.net/rss/kinh-doanh.rss'),('Công Nghệ','https://vnexpress.net/rss/so-hoa.rss'),('Thể Thao','https://vnexpress.net/rss/the-thao.rss'),('Giải Trí','https://vnexpress.net/rss/giai-tri.rss'),('Sức Khỏe','https://vnexpress.net/rss/suc-khoe.rss'),('Pháp Luật','https://vnexpress.net/rss/phap-luat.rss')]
+    arts=[]
+    with ThreadPoolExecutor(6) as ex:
+        futs=[ex.submit(_rss_articles,u,g,'vne',6) for g,u in feeds]
+        for f in as_completed(futs,timeout=8):
+            try:arts.extend(f.result() or [])
+            except Exception:pass
+    return arts
+
+def _fetch_shorts():
+    from main import SHORTS_FALLBACK, _yt_channel_shorts
+    items=[];seen=set()
+    for ch in ["baodantri7941","baosuckhoedoisongboyte"]:
+        got=_yt_channel_shorts(ch,20)
+        for v in got:
+            vid=v.get('id')
+            if vid and vid not in seen:seen.add(vid);items.append(v)
+    for v in SHORTS_FALLBACK:
+        vid=v.get('id')
+        if vid and vid not in seen:seen.add(vid);items.append(v)
+    return items[:40]
+
+def _bg_refresh():
+    try:
+        data=_fetch_homepage()
+        if data:_HOME_CACHE.update({"t":time.time(),"d":data})
+    except Exception:pass
+    try:
+        data=_fetch_shorts()
+        if data:_SHORTS_CACHE.update({"t":time.time(),"d":data})
+    except Exception:pass
+
+@app.on_event("startup")
+async def startup():threading.Thread(target=_bg_refresh,daemon=True).start()
+
+# Periodic refresh every 8 min.
+def _periodic():
+    while True:
+        time.sleep(480);_bg_refresh()
+threading.Thread(target=_periodic,daemon=True).start()
+
+# ===== OVERRIDE HOMEPAGE/SHORTS =====
+_rm={'/api/homepage','/api/shorts','/api/topic_post','/api/article/ask','/'}
+app.router.routes=[r for r in app.router.routes if getattr(r,'path',None) not in _rm]
+
+@app.get('/api/homepage')
+def api_homepage():
+    if _HOME_CACHE['d']:
+        if time.time()-_HOME_CACHE['t']>300:threading.Thread(target=_bg_refresh,daemon=True).start()
+        return JSONResponse(_HOME_CACHE['d'])
+    data=_fetch_homepage()
+    if data:_HOME_CACHE.update({"t":time.time(),"d":data})
+    return JSONResponse(data)
+
+@app.get('/api/shorts')
+def api_shorts(refresh:int=Query(default=0)):
+    if _SHORTS_CACHE['d'] and not refresh:return JSONResponse(_SHORTS_CACHE['d'])
+    if _SHORTS_CACHE['d'] and time.time()-_SHORTS_CACHE['t']<120:return JSONResponse(_SHORTS_CACHE['d'])
+    data=_fetch_shorts()
+    if data:_SHORTS_CACHE.update({"t":time.time(),"d":data})
+    return JSONResponse(data or _SHORTS_CACHE.get('d',[]))
+
+# ===== TOPIC POST (no prompt leaking, source_details persistent) =====
+def _fast_rss_sources(topic,limit=8):
+    feeds=[('VnExpress','https://vnexpress.net/rss/tin-moi-nhat.rss'),('Dân trí','https://dantri.com.vn/rss/home.rss'),('Vietnamnet','https://vietnamnet.vn/rss/tin-moi-nhat.rss')]
+    pool=[]
+    for name,url in feeds:
+        try:
+            r=requests.get(url,headers=UA,timeout=5);r.encoding='utf-8';soup=BeautifulSoup(r.text,'xml')
+            for it in soup.find_all('item')[:20]:
+                title=clean(it.find('title').get_text(' ',strip=True) if it.find('title') else '')
+                link=clean(it.find('link').get_text(strip=True) if it.find('link') else '')
+                desc=clean(BeautifulSoup(it.find('description').get_text(' ',strip=True) if it.find('description') else '','lxml').get_text(' ',strip=True))
+                # Extract image from description HTML
+                desc_html=it.find('description').get_text(' ',strip=True) if it.find('description') else ''
+                img_match=re.search(r'<img[^>]+src=["\']([^"\']+)',desc_html)
+                img=img_match.group(1) if img_match else ''
+                if title and link:pool.append({'title':title,'url':link,'source':name,'snippet':desc,'img':img})
+        except Exception:pass
+    # Score by topic relevance
+    keys=[w.lower() for w in re.findall(r'[A-Za-zÀ-ỹ0-9]+',topic) if len(w)>2]
+    scored=[]
+    for it in pool:
+        hay=(it['title']+' '+it['snippet']).lower()
+        sc=sum(2 for k in keys if k in hay)
+        if topic.lower() in hay:sc+=10
+        if sc>0:scored.append((sc,it))
+    scored.sort(key=lambda x:x[0],reverse=True)
+    return [it for _,it in scored[:limit]]
+
+@app.post('/api/topic_post')
+async def topic_post(request:Request):
+    body=await request.json();topic=clean(body.get('topic',''))
+    if not topic:return JSONResponse({'error':'missing topic'},status_code=400)
+    img=pollinations_image_url(topic) if pollinations_image_url else ''
+    sources=_fast_rss_sources(topic,8)
+    if not sources:return JSONResponse({'error':'Không tìm được nội dung liên quan. Thử chủ đề khác.'},status_code=422)
+    ctx='\n\n'.join([f"Nguồn: {s['source']}\nTiêu đề: {s['title']}\nTóm tắt: {s['snippet']}" for s in sources[:6]])
+    prompt=f"""Viết bài tiếng Việt về: {topic}
+
+Dữ liệu nguồn:
+{ctx[:12000]}
+
+Yêu cầu: Chỉ viết về {topic}. Bỏ mọi nội dung không liên quan. Tiêu đề hấp dẫn. Sapo 2 câu. 5-7 đoạn phân tích. Cuối ghi nguồn tham khảo."""
+    text=None
+    if qwen_generate:
+        try:text=await asyncio.wait_for(qwen_generate(prompt,image_url=img,max_tokens=1500),timeout=30)
+        except Exception:pass
+    if not text or len(text)<300:
+        bullets='\n'.join([f"• {s['title']}" for s in sources[:6]])
+        vias=', '.join(sorted({s['source'] for s in sources}))
+        text=f"{topic}\n\n{bullets}\n\nNguồn tham khảo: {vias}"
+    # Remove any leaked prompt/rules from output.
+    for marker in ['QUY TẮC BẮT BUỘC','NỘI DUNG NGUỒN:','Dữ liệu nguồn:','Yêu cầu:','Viết bài tiếng Việt về:']:
+        if marker in text:text=text.split(marker)[0].strip()
+    post=make_post(topic,text,img,'','topic',sources=[{'title':s['title'],'url':s['url'],'via':s['source'],'img':s.get('img','')} for s in sources])
+    post['images']=[img]
+    post['source_details']=[{'title':s['title'],'url':s['url'],'via':s['source'],'img':s.get('img',''),'content':s['snippet'][:500]} for s in sources]
+    posts=_load_ai_wall();posts.insert(0,post);_save_ai_wall(posts)
+    return JSONResponse({'post':post})
+
+# ===== ASK AI (uses article/short content as context) =====
+@app.post('/api/article/ask')
+async def article_ask(request:Request):
+    body=await request.json();url=clean(body.get('url',''));question=clean(body.get('question',''));context=clean(body.get('context',''))
+    if not question:return JSONResponse({'error':'missing question'},status_code=400)
+    if url:
+        try:
+            data=scrape_any_url(url) if scrape_any_url else None
+            if data:context=(data.get('title','')+'\n'+data.get('summary','')+'\n'+data.get('text','')).strip()
+        except Exception:pass
+    if not context:context=question
+    prompt=f"""Trả lời chi tiết câu hỏi dựa trên nội dung bài viết/video bên dưới. Chỉ dựa vào nội dung có sẵn.
+
+Nội dung: {context[:8000]}
+
+Câu hỏi: {question}"""
+    ans=None
+    if qwen_generate:
+        try:ans=await asyncio.wait_for(qwen_generate(prompt,max_tokens=800),timeout=25)
+        except Exception:pass
+    if not ans:ans='AI chưa trả lời được lúc này.'
+    return JSONResponse({'answer':ans})
+
+# ===== ROOT: load UI + inject =====
+def _load_index():
+    try:
+        with open('/app/static/index.html','r',encoding='utf-8') as f:html=f.read()
+    except Exception:html=''
+    if '<!DOCTYPE html>' not in html or '<div id="view-home"' not in html:
+        try:
+            r=requests.get(RESTORE_INDEX_URL,timeout=15)
+            if r.status_code==200 and '<!DOCTYPE html>' in r.text:html=r.text
+        except Exception:pass
+    return html or '<!DOCTYPE html><html><body>VNEWS</body></html>'
+
+INJECT=r'''
+<style>
+.ai-compose{width:calc(100% - 8px)!important}.ai-compose-row{display:flex!important;flex-direction:column!important;gap:8px!important}.ai-compose-row input,.ai-compose-row button{width:100%!important;box-sizing:border-box!important}.article-ai-ask{margin-top:12px;background:#141414;border:1px solid #2a2a2a;border-radius:10px;padding:10px}.article-ai-ask textarea{width:100%;min-height:60px;background:#222;border:1px solid #444;color:#eee;border-radius:8px;padding:8px;font-size:12px}.article-ai-ask button{background:#2d8659;border:0;color:#fff;border-radius:8px;padding:7px 12px;margin-top:6px;font-size:11px}.article-ai-answer{white-space:pre-wrap;color:#ccc;font-size:12px;line-height:1.5;margin-top:8px}.source-in-topic{background:#1a1a1a;border:1px solid #2a2a2a;border-radius:8px;padding:8px;margin:6px 0;cursor:pointer}.source-in-topic img{width:100%;aspect-ratio:16/9;object-fit:cover;border-radius:6px;margin-bottom:6px;background:#222}.source-in-topic .src-title{font-size:12px;color:#5cb87a;font-weight:700}.source-in-topic .src-via{font-size:10px;color:#888}
+</style>
+<script>
+(function(){
+function esc(s){return String(s||'').replace(/[&<>"']/g,m=>({'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;',"'":'&#39;'}[m]));}
+// Patch readArticle to add Ask AI box + single rewrite button (no duplicate).
+let _origRead=window.readArticle;
+if(_origRead){window.readArticle=async function(){let ret=await _origRead.apply(this,arguments);setTimeout(()=>{let art=document.querySelector('#view-article .article-view');if(!art)return;
+// Add Ask AI box if not present.
+if(!document.getElementById('article-ai-ask')){let box=document.createElement('div');box.id='article-ai-ask';box.className='article-ai-ask';box.innerHTML='<b style="color:#5cb87a;font-size:13px">🤖 Hỏi AI về bài viết</b><textarea id="article-ai-q" placeholder="Nhập câu hỏi..."></textarea><button onclick="askArticleAI()">Hỏi</button><div id="article-ai-ans" class="article-ai-answer"></div>';art.appendChild(box);}
+// Add rewrite button only if not already present.
+let actions=art.querySelector('.article-actions');if(actions&&!actions.querySelector('[data-rw]')){let btn=document.createElement('button');btn.className='primary';btn.setAttribute('data-rw','1');btn.textContent='🤖 AI viết lại & đăng tường';btn.onclick=function(){if(typeof rewriteCurrentArticle==='function')rewriteCurrentArticle();};actions.appendChild(btn);}},500);return ret;}}
+window.askArticleAI=async function(){let q=document.getElementById('article-ai-q')?.value.trim();if(!q)return;let ans=document.getElementById('article-ai-ans');ans.textContent='Đang hỏi...';let url=(window._currentArticle?.url)||'';let ctx=document.querySelector('.article-view')?.innerText?.slice(0,6000)||'';let r=await fetch('/api/article/ask',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({url,question:q,context:ctx})}).then(r=>r.json()).catch(()=>({answer:'Lỗi'}));ans.textContent=r.answer||'Không có trả lời';}
+// Patch topic wall read to show source_details with images and in-app links.
+let _origWallRead=window.readLiveTopicWall||window.aiReadWallPatched||window.aiReadWall;
+function renderTopicArticle(p){
+showView('view-article');
+let srcHtml='';
+if(p.source_details&&p.source_details.length){srcHtml='<div style="margin-top:12px"><b style="color:#5cb87a;font-size:13px">📚 Bài viết liên quan</b>';p.source_details.forEach(s=>{srcHtml+=`<div class="source-in-topic" onclick="readArticle('${esc(s.url)}')">${s.img?`<img src="${esc(s.img)}" onerror="this.style.display='none'">`:''}<div class="src-title">${esc(s.title)}</div><div class="src-via">${esc(s.via||'')}</div></div>`;});srcHtml+='</div>';}
+let h=`<button class="back-btn" onclick="switchCat('home')">← Quay lại</button><div class="article-view"><span class="badge badge-ai">AI</span><h1 class="article-title">${esc(p.title)}</h1>${p.img?`<img class="article-img" src="${esc(p.img)}">`:''}<p class="article-p" style="white-space:pre-wrap">${esc(p.text)}</p>${srcHtml}<div class="article-actions"><button onclick="doShare('${esc(p.title)}','${esc(p.url||location.href)}','${esc(p.img||'')}')">📤 Chia sẻ</button></div><div class="article-ai-ask" id="article-ai-ask"><b style="color:#5cb87a;font-size:13px">🤖 Hỏi AI về bài này</b><textarea id="article-ai-q" placeholder="Hỏi bất cứ điều gì..."></textarea><button onclick="askArticleAI()">Hỏi</button><div id="article-ai-ans" class="article-ai-answer"></div></div></div>`;
+document.getElementById('view-article').innerHTML=h;window.scrollTo(0,0);
+// Store context for Ask AI.
+window._currentArticle={url:'',data:{title:p.title,text:p.text}};
+}
+window.readLiveTopicWall=window.aiReadWallPatched=window.aiReadWall=function(i){let arr=window.__topicWallE||window.liveTopicWall||window.patchedWall||window.finalWall||window.finalWall3||[];let p=arr[i];if(p){renderTopicArticle(p);return;}if(_origWallRead)_origWallRead(i);};
+// Override Ask AI to use stored article text as context when no URL.
+let _origAsk=window.askArticleAI;
+window.askArticleAI=async function(){let q=document.getElementById('article-ai-q')?.value.trim();if(!q)return;let ans=document.getElementById('article-ai-ans');ans.textContent='Đang hỏi AI...';let url=window._currentArticle?.url||'';let ctx=window._currentArticle?.data?.text||document.querySelector('.article-view')?.innerText?.slice(0,6000)||'';let r=await fetch('/api/article/ask',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({url,question:q,context:ctx})}).then(r=>r.json()).catch(()=>({answer:'Lỗi kết nối'}));ans.textContent=r.answer||'Không trả lời được';}
+})();
+</script>
+'''
+
+@app.get('/')
+async def index():
+    html=_load_index()
+    # Only inject our lightweight patch — no 6-layer chain.
+    from ai_patch import PATCH_INJECT as OLD_INJECT
+    return HTMLResponse(html.replace('</body>',OLD_INJECT+INJECT+'\n</body>') if '</body>' in html else html+INJECT)