Delete patch_runtime.py, restore_runner.py, rewrite_slide.py

patch_runtime.py

@@ -1,274 +0,0 @@
-"""Runtime patch layer for VNEWS.
-Keeps the current large app intact, but replaces fragile AI wall endpoints with
-stable JSON endpoints and injects frontend safeJson wrappers.
-"""
-import hashlib
-import time
-import os
-from urllib.parse import quote
-
-import requests
-from bs4 import BeautifulSoup
-from fastapi import Request
-from fastapi.responses import JSONResponse, HTMLResponse
-
-import main as _main
-
-app = _main.app
-DEFAULT_IMG = "https://s1.vnecdn.net/vnexpress/restruct/i/v9505/logo_default.jpg"
-
-
-def _remove_routes(paths):
-    app.router.routes = [r for r in app.router.routes if getattr(r, "path", None) not in set(paths)]
-
-
-def _safe_text(v):
-    return (v or "").strip()
-
-
-def _ensure_article(url: str):
-    data = None
-    try:
-        if hasattr(_main, "_article_by_url"):
-            data = _main._article_by_url(url)
-    except Exception:
-        data = None
-    if not data:
-        try:
-            data = _main._scrape_generic_article(url) if hasattr(_main, "_scrape_generic_article") else None
-        except Exception:
-            data = None
-    if not data:
-        data = {"title": "", "summary": "", "og_image": "", "body": [], "url": url, "source": "generic"}
-    title = _safe_text(data.get("title"))
-    summary = _safe_text(data.get("summary"))
-    img = _safe_text(data.get("og_image"))
-    body = data.get("body") or []
-    if not title or not summary or not img or not body:
-        try:
-            r = requests.get(url, headers=getattr(_main, "HEADERS", {}), timeout=15)
-            r.encoding = "utf-8"
-            soup = BeautifulSoup(r.text, "lxml")
-            if not title:
-                tag = soup.find("meta", property="og:title") or soup.find("title")
-                title = tag.get("content", "").strip() if tag and tag.name == "meta" else (tag.get_text(strip=True) if tag else "")
-            if not summary:
-                tag = soup.find("meta", property="og:description") or soup.find("meta", attrs={"name": "description"})
-                summary = tag.get("content", "").strip() if tag else ""
-            if not img:
-                tag = soup.find("meta", property="og:image") or soup.find("meta", attrs={"name": "twitter:image"})
-                img = tag.get("content", "").strip() if tag else ""
-            if not body:
-                ps = []
-                for p in soup.find_all("p"):
-                    t = p.get_text(" ", strip=True)
-                    if len(t) > 40:
-                        ps.append({"type": "p", "text": t})
-                    if len(ps) >= 30:
-                        break
-                body = ps
-        except Exception:
-            pass
-    if not summary and body:
-        first = next((b.get("text", "") for b in body if b.get("type") == "p" and b.get("text")), "")
-        summary = first[:360]
-    if not title:
-        title = url
-    if not img:
-        img = DEFAULT_IMG
-    if not body and summary:
-        body = [{"type": "p", "text": summary}]
-    data.update({"title": title, "summary": summary, "og_image": img, "body": body, "url": url})
-    return data
-
-
-def _rewrite(data, tone="tu-nhien"):
-    try:
-        if hasattr(_main, "_ai_rewrite_article"):
-            text = _main._ai_rewrite_article(data, tone=tone)
-            if text and len(text.strip()) > 50:
-                return text.strip()
-    except Exception:
-        pass
-    title = data.get("title", "")
-    summary = data.get("summary", "")
-    ps = [b.get("text", "") for b in data.get("body", []) if b.get("type") == "p" and b.get("text")]
-    lead = summary or (ps[0] if ps else "")
-    points = "\n".join(["• " + p[:220] + ("..." if len(p) > 220 else "") for p in ps[:5]])
-    body = "\n\n".join(ps[:10])
-    return (f"Bản tin AI viết lại: {title}\n\n{lead}\n\n{body}\n\nĐiểm chính:\n{points}").strip()
-
-
-def _topic_image(topic):
-    try:
-        if hasattr(_main, "_image_for_topic"):
-            return _main._image_for_topic(topic)
-    except Exception:
-        pass
-    return "https://image.pollinations.ai/prompt/" + quote("editorial illustration Vietnamese news " + topic, safe="") + "?width=1024&height=576&nologo=true"
-
-
-def _save_post(post):
-    try:
-        posts = _main._load_wall() if hasattr(_main, "_load_wall") else []
-    except Exception:
-        posts = []
-    posts.insert(0, post)
-    try:
-        if hasattr(_main, "_save_wall"):
-            _main._save_wall(posts)
-    except Exception:
-        pass
-    return post
-
-
-_remove_routes(["/api/url_wall", "/api/topic_post", "/api/rewrite_share", "/"])
-
-
-@app.post("/api/url_wall")
-async def patched_url_wall(request: Request):
-    try:
-        body = await request.json()
-    except Exception:
-        body = {}
-    url = _safe_text(body.get("url"))
-    tone = _safe_text(body.get("tone")) or "tu-nhien"
-    if not url:
-        return JSONResponse({"error": "missing url"}, status_code=400)
-    try:
-        data = _ensure_article(url)
-        text = _rewrite(data, tone=tone)
-        post = {
-            "id": hashlib.md5((url + str(time.time())).encode()).hexdigest()[:12],
-            "url": url,
-            "title": data.get("title") or url,
-            "summary": data.get("summary") or "",
-            "img": data.get("og_image") or DEFAULT_IMG,
-            "text": text or (data.get("summary") or data.get("title") or url),
-            "source": data.get("source", "url"),
-            "ts": int(time.time()),
-        }
-        _save_post(post)
-        return JSONResponse({"post": post})
-    except Exception as e:
-        return JSONResponse({"error": "Không tạo được tóm tắt URL", "detail": str(e)[:300]}, status_code=500)
-
-
-@app.post("/api/rewrite_share")
-async def patched_rewrite_share(request: Request):
-    return await patched_url_wall(request)
-
-
-@app.post("/api/topic_post")
-async def patched_topic_post(request: Request):
-    try:
-        body = await request.json()
-    except Exception:
-        body = {}
-    topic = _safe_text(body.get("topic"))
-    tone = _safe_text(body.get("tone")) or "tu-nhien"
-    if not topic:
-        return JSONResponse({"error": "missing topic"}, status_code=400)
-    try:
-        context = ""
-        try:
-            if hasattr(_main, "_topic_article_context"):
-                context = _main._topic_article_context(topic)
-            if not context and hasattr(_main, "_web_context"):
-                context = _main._web_context(topic)
-        except Exception:
-            context = ""
-        if not context:
-            context = f"Chủ đề: {topic}"
-        data = {"title": topic, "summary": context[:420], "og_image": _topic_image(topic), "body": [{"type": "p", "text": context}], "source": "topic", "url": ""}
-        text = _rewrite(data, tone=tone)
-        post = {
-            "id": hashlib.md5((topic + str(time.time())).encode()).hexdigest()[:12],
-            "url": "",
-            "title": topic,
-            "summary": data["summary"],
-            "img": data["og_image"] or DEFAULT_IMG,
-            "text": text or context,
-            "source": "topic",
-            "ts": int(time.time()),
-        }
-        _save_post(post)
-        return JSONResponse({"post": post})
-    except Exception as e:
-        return JSONResponse({"error": "Không tạo được bài theo chủ đề", "detail": str(e)[:300]}, status_code=500)
-
-
-_FRONTEND_PATCH = r'''
-<script>
-(function(){
-  async function safeJson(res){
-    const text = await res.text();
-    try { return JSON.parse(text); }
-    catch(e){ return { error: (text || 'Server không trả JSON').slice(0,500) }; }
-  }
-  window.safeJson = safeJson;
-  window.createUrlPost = function(){
-    let inp=document.getElementById('ai-url-input');
-    let url=(inp&&inp.value||'').trim();
-    if(!url){ alert('Dán URL trước'); return; }
-    fetch('/api/url_wall',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({url})})
-      .then(safeJson).then(j=>{
-        if(j&&j.post){
-          if(!j.post.img) j.post.img='https://s1.vnecdn.net/vnexpress/restruct/i/v9505/logo_default.jpg';
-          if(!j.post.text) j.post.text=j.post.summary||j.post.title||'Không lấy được nội dung tóm tắt.';
-          if(typeof prependWallPost==='function') prependWallPost(j.post);
-          alert('Đã tóm tắt URL và đăng lên tường');
-          if(inp) inp.value='';
-        } else alert((j&&j.error)||'Lỗi URL');
-      }).catch(e=>alert('Lỗi URL: '+e.message));
-  };
-  window.createTopicPost = function(){
-    let inp=document.getElementById('ai-topic-input');
-    let topic=(inp&&inp.value||'').trim();
-    if(!topic){ alert('Nhập chủ đề trước'); return; }
-    fetch('/api/topic_post',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({topic})})
-      .then(safeJson).then(j=>{
-        if(j&&j.post){
-          if(!j.post.img) j.post.img='https://s1.vnecdn.net/vnexpress/restruct/i/v9505/logo_default.jpg';
-          if(!j.post.text) j.post.text=j.post.summary||j.post.title||'Không lấy được nội dung.';
-          if(typeof prependWallPost==='function') prependWallPost(j.post);
-          alert('Đã tạo bài và đăng lên tường');
-          if(inp) inp.value='';
-        } else alert((j&&j.error)||'Lỗi tạo bài');
-      }).catch(e=>alert('Lỗi tạo bài: '+e.message));
-  };
-  window.rewriteCurrentArticle = function(){
-    if(!window._currentArticle && typeof _currentArticle!=='undefined') window._currentArticle=_currentArticle;
-    let ca = (typeof _currentArticle!=='undefined') ? _currentArticle : window._currentArticle;
-    if(!ca || !ca.url){ alert('Chưa có bài viết để rewrite'); return; }
-    let tone=document.getElementById('rewrite-tone')?.value||'nghiem-tuc';
-    let btn=document.querySelector('.article-actions button.primary');
-    if(btn){btn.textContent='Đang rewrite...';btn.disabled=true;}
-    fetch('/api/rewrite_share',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({url:ca.url,tone})})
-      .then(safeJson).then(j=>{
-        if(j&&j.post){
-          if(!j.post.img) j.post.img='https://s1.vnecdn.net/vnexpress/restruct/i/v9505/logo_default.jpg';
-          if(!j.post.text) j.post.text=j.post.summary||j.post.title||'Không lấy được nội dung.';
-          let box=document.getElementById('rewrite-result');
-          if(box) box.innerHTML='<div class="rewrite-box"><div class="rewrite-title">Đã rewrite và đăng lên Tường AI</div><div class="rewrite-text">'+(j.post.text||'').replace(/[&<>"']/g,m=>({'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;',"'":'&#39;'}[m]))+'</div></div>';
-          if(typeof prependWallPost==='function') prependWallPost(j.post);
-          alert('Đã đăng lên Tường AI');
-        } else alert((j&&j.error)||'Không tạo được bài AI');
-      }).catch(e=>alert('Lỗi tạo bài AI: '+e.message))
-      .finally(()=>{if(btn){btn.textContent='🤖 AI viết lại & đăng tường';btn.disabled=false;}});
-  };
-})();
-</script>
-'''
-
-
-@app.get("/")
-async def patched_index():
-    try:
-        with open("/app/static/index.html", "r", encoding="utf-8") as f:
-            html = f.read()
-        if "window.safeJson" not in html:
-            html = html.replace("</body>", _FRONTEND_PATCH + "</body>")
-        return HTMLResponse(content=html)
-    except Exception as e:
-        return HTMLResponse(content=f"<pre>Index error: {str(e)}</pre>", status_code=500)

restore_runner.py

@@ -1,31 +0,0 @@
-import os
-import sys
-import subprocess
-from huggingface_hub import snapshot_download
-
-REVISION = os.environ.get("VNEWS_RESTORE_REVISION", "bcaa2dc")
-REPO_ID = os.environ.get("VNEWS_REPO_ID", "bep40/vnews")
-
-# Download exact Space snapshot from Hugging Face Hub.
-# This avoids manually copying huge files from an old commit.
-snapshot_dir = snapshot_download(
-    repo_id=REPO_ID,
-    repo_type="space",
-    revision=REVISION,
-    local_dir="/tmp/vnews_restore",
-    local_dir_use_symlinks=False,
-)
-
-os.chdir(snapshot_dir)
-sys.path.insert(0, snapshot_dir)
-
-# Commit bcaa2dc Dockerfile ran ai_patch:app.
-cmd = [
-    "uvicorn",
-    "ai_patch:app",
-    "--host",
-    "0.0.0.0",
-    "--port",
-    "7860",
-]
-os.execvp(cmd[0], cmd)

rewrite_slide.py

@@ -1,185 +0,0 @@
-"""Fast rewrite as slides - no AI needed, extracts key points + images from article."""
-from main import app
-from fastapi import Request
-from fastapi.responses import JSONResponse
-import requests, re, time, random, json, os
-from bs4 import BeautifulSoup
-from urllib.parse import quote
-
-UA = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36', 'Accept-Language': 'vi-VN,vi;q=0.9'}
-
-try:
-    from main import _load_wall, _save_wall
-except:
-    _data_dir = "/data" if os.path.isdir("/data") else "/app/data"
-    _wall_file = os.path.join(_data_dir, "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)
-            with open(_wall_file+'.tmp', 'w', encoding='utf-8') as f: json.dump(posts[:100], f, ensure_ascii=False)
-            os.replace(_wall_file+'.tmp', _wall_file)
-        except: pass
-
-
-def _clean(s): return re.sub(r'\s+', ' ', str(s or '')).strip()
-
-
-def _scrape_article_full(url):
-    """Scrape article: extract paragraphs + ALL images."""
-    try:
-        r = requests.get(url, headers=UA, timeout=15, allow_redirects=True)
-        r.encoding = 'utf-8'
-        soup = BeautifulSoup(r.text, 'lxml')
-        for tag in soup.find_all(['script', 'style', 'nav', 'footer', 'aside', 'form']): tag.decompose()
-        
-        # Title
-        h1 = soup.find('h1')
-        ogt = soup.find('meta', property='og:title')
-        title = (h1.get_text(strip=True) if h1 else '') or (ogt.get('content', '') if ogt else '')
-        
-        # OG image
-        ogi = soup.find('meta', property='og:image')
-        og_img = ogi.get('content', '') if ogi else ''
-        if og_img and og_img.startswith('//'): og_img = 'https:' + og_img
-        
-        # Find content block
-        block = None
-        for sel in ['article', '.singular-content', '.detail-content', '.fck_detail', '.content-detail', '.knc-content', 'main', '.cms-body', '.article__body']:
-            el = soup.select_one(sel)
-            if el and len(el.find_all('p')) >= 2: block = el; break
-        if not block: block = soup.body or soup
-        
-        # Extract paragraphs and images IN ORDER
-        paragraphs = []
-        images = []
-        seen_imgs = set()
-        
-        if og_img and og_img not in seen_imgs:
-            images.append(og_img)
-            seen_imgs.add(og_img)
-        
-        for el in block.find_all(['p', 'h2', 'h3', 'figure', 'img'], recursive=True):
-            if el.name == 'p':
-                t = _clean(el.get_text(strip=True))
-                if t and len(t) > 40:
-                    paragraphs.append(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') or ''
-                    if src and 'base64' not in src:
-                        if src.startswith('//'): src = 'https:' + src
-                        if src not in seen_imgs:
-                            images.append(src)
-                            seen_imgs.add(src)
-        
-        return {'title': _clean(title), 'paragraphs': paragraphs, 'images': images, 'og_img': og_img}
-    except Exception as e:
-        return None
-
-
-def _extract_key_points(paragraphs, max_points=5):
-    """Extract key points: take first sentence of each significant paragraph."""
-    points = []
-    for p in paragraphs:
-        if len(points) >= max_points: break
-        # Take first complete sentence (ends with . ! ?)
-        m = re.match(r'^(.+?[.!?])\s', p)
-        if m:
-            sentence = m.group(1)
-        else:
-            sentence = p[:150] + ('.' if not p.endswith('.') else '')
-        
-        # Skip if too short or duplicate
-        if len(sentence) < 30: continue
-        if any(sentence[:50] in existing for existing in points): continue
-        
-        points.append(sentence)
-    
-    return points
-
-
-@app.post("/api/rewrite_slide")
-async def api_rewrite_slide(request: Request):
-    """
-    Fast rewrite as SLIDES:
-    - Extract key points from article (1 sentence each, full and complete)
-    - Pair each point with an image from the article
-    - Return as slides array for frontend to display
-    - Save to Tường AI
-    NO AI NEEDED - instant response.
-    """
-    body = await request.json()
-    url = _clean(body.get("url", ""))
-    context = body.get("context", "")
-    
-    if not url and not context:
-        return JSONResponse({"error": "Cần URL hoặc nội dung"}, status_code=400)
-    
-    # Scrape article
-    data = None
-    if url and url.startswith("http"):
-        data = _scrape_article_full(url)
-    
-    if not data and context:
-        # Use context passed from frontend
-        paragraphs = [_clean(p) for p in context.split('\n') if len(_clean(p)) > 40]
-        data = {'title': paragraphs[0][:80] if paragraphs else 'Bài viết', 'paragraphs': paragraphs, 'images': [], 'og_img': ''}
-    
-    if not data or not data.get('paragraphs'):
-        return JSONResponse({"error": "Không đọc được bài viết"}, status_code=422)
-    
-    # Extract key points
-    points = _extract_key_points(data['paragraphs'], max_points=6)
-    if not points:
-        return JSONResponse({"error": "Không tìm được ý chính"}, status_code=422)
-    
-    # Build slides: pair each point with an image
-    images = data.get('images', [])
-    slides = []
-    for i, point in enumerate(points):
-        img = images[i] if i < len(images) else (images[-1] if images else '')
-        # Proxy dantri images
-        if img and 'cdnphoto.dantri' in img:
-            img = '/api/proxy/img?url=' + quote(img, safe='')
-        slides.append({
-            'text': point,
-            'image': img,
-            'index': i + 1
-        })
-    
-    # Create post for Tường AI
-    summary_text = '\n\n'.join([f"• {s['text']}" for s in slides])
-    # Auto voice + emotion based on topic (reuse ai_ext detector if available)
-    try:
-        from ai_ext import _detect_voice_emotion
-        _voice, _emotion = _detect_voice_emotion(data['title'], summary_text)
-    except Exception:
-        _voice, _emotion = "hoaimy", "trung_tinh"
-    post = {
-        "id": str(int(time.time() * 1000)) + str(random.randint(100, 999)),
-        "title": data['title'],
-        "text": summary_text,
-        "img": images[0] if images else '',
-        "url": url,
-        "kind": "slide_summary",
-        "slides": slides,
-        "images": images[:10],
-        "video": "",
-        "voice": _voice,
-        "emotion": _emotion,
-        "ts": int(time.time())
-    }
-    
-    # Save to wall
-    posts = _load_wall()
-    posts.insert(0, post)
-    _save_wall(posts)
-    
-    return JSONResponse({"post": post, "slides": slides})