Sostituita ricerca web SearXNG con Chromium/Playwright; aggiunto Dockerfile per Chromium su HF Space

Dockerfile

@@ -0,0 +1,27 @@
+FROM python:3.11-slim
+
+WORKDIR /app
+
+# Dipendenze di sistema richieste da Chromium (Playwright)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    wget gnupg ca-certificates \
+    libnss3 libnspr4 libatk1.0-0 libatk-bridge2.0-0 libcups2 \
+    libdrm2 libdbus-1-3 libxkbcommon0 libxcomposite1 libxdamage1 \
+    libxfixes3 libxrandr2 libgbm1 libasound2 libpango-1.0-0 \
+    libcairo2 libatspi2.0-0 libx11-6 libxext6 fonts-liberation \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+RUN python -m playwright install chromium
+
+COPY . .
+
+# HF Spaces esegue il container come utente non-root (id 1000)
+RUN mkdir -p /app/database && chmod -R 777 /app
+USER 1000
+
+EXPOSE 7860
+ENV PYTHONUNBUFFERED=1
+
+CMD ["python", "app.py"]

app.py

@@ -24,7 +24,6 @@ def page(title: str, content: str, active: str = "") -> str:
     nav_items = [
         ("/",            "🏠", "Home"),
         ("/chat",        "💬", "Chat"),
-        ("/instances/ui","🌐", "Istanze"),
         ("/commands",    "⚡", "Comandi"),
         ("/docs",        "📖", "API Docs"),
     ]
@@ -101,7 +100,7 @@ async def home():
     <tr><th>Componente</th><th>Stato</th><th>Info</th></tr>
     <tr><td>AI Brain</td><td><span class="badge {badge}">{status}</span></td><td>KB + Web search</td></tr>
     <tr><td>Knowledge Base</td><td><span class="badge green">Online</span></td><td>{kb} documenti in memoria</td></tr>
-    <tr><td>Ricerca Web</td><td><span class="badge blue">SearXNG</span></td><td>Multi-istanza con fallback</td></tr>
+    <tr><td>Ricerca Web</td><td><span class="badge blue">Chromium</span></td><td>Browser headless (Playwright)</td></tr>
   </table>
 </div>
 
@@ -109,7 +108,6 @@ async def home():
   <h2>Accesso rapido</h2>
   <div style="display:flex;gap:10px;flex-wrap:wrap">
     <a href="/chat"         class="btn btn-primary">💬 Chat</a>
-    <a href="/instances/ui" class="btn btn-secondary">🌐 Istanze SearXNG</a>
     <a href="/commands"     class="btn btn-secondary">⚡ Comandi terminale</a>
     <a href="/docs"         class="btn btn-secondary">📖 API Docs</a>
   </div>
@@ -191,91 +189,6 @@ document.getElementById('inp').addEventListener('keydown', e => {{
     return HTMLResponse(page("Chat", content, "/chat"))
 
 
-# ── Istanze UI ─────────────────────────────────────────────────────────────────
-
-@app.get("/instances/ui", response_class=HTMLResponse)
-def instances_ui():
-    content = f"""
-<h1>🌐 Istanze SearXNG</h1>
-<p class="sub">Testa tutte le istanze pubbliche e seleziona la più veloce.</p>
-
-<div class="card">
-  <div style="display:flex;gap:10px;align-items:center;flex-wrap:wrap">
-    <button class="btn btn-primary" onclick="testAll()">🔍 Testa tutte</button>
-    <span id="scan-status" style="color:#888;font-size:13px"></span>
-  </div>
-  <div id="active-info" style="margin-top:12px;font-size:13px;color:#888"></div>
-</div>
-
-<div id="results-card" class="card" style="display:none">
-  <h2>Risultati <span id="count-badge"></span></h2>
-  <table>
-    <thead><tr><th>Istanza</th><th>Stato</th><th>Velocità</th><th>Paese</th><th>Versione</th><th>Uptime</th><th>HTTPS</th><th>Motori</th><th>Azione</th></tr></thead>
-    <tbody id="inst-table"></tbody>
-  </table>
-</div>
-
-<script>
-async function loadActive() {{
-  const r = await fetch('{BASE}/instances/active');
-  const d = await r.json();
-  document.getElementById('active-info').innerHTML =
-    'Istanza attiva: <span style="color:#7c9ef8;font-family:monospace">' + d.active + '</span>';
-}}
-
-async function select(url) {{
-  const r = await fetch('{BASE}/instances/select?url=' + encodeURIComponent(url), {{method:'POST'}});
-  const d = await r.json();
-  if (d.ok) {{ alert('Impostata: ' + d.active); loadActive(); }}
-  else alert('Errore: ' + JSON.stringify(d));
-}}
-
-async function testAll() {{
-  document.getElementById('scan-status').innerText = 'Scansione in corso... (~30s)';
-  document.getElementById('results-card').style.display = 'none';
-  const r = await fetch('{BASE}/instances');
-  const d = await r.json();
-  document.getElementById('scan-status').innerText =
-    'Completato: ' + d.online_with_results + ' online su ' + d.total_tested + ' testate';
-  document.getElementById('count-badge').innerHTML =
-    '<span class="badge green">' + d.online_with_results + ' online</span>';
-
-  const tbody = document.getElementById('inst-table');
-  tbody.innerHTML = '';
-  d.all.forEach(i => {{
-    const ok = i.online && i.results;
-    const color = ok ? '#3cb371' : i.online ? '#f0a000' : '#e05555';
-    const stato = ok ? '✅ Online' : i.online ? '⚠️ No risultati' : '❌ Offline';
-    const ms    = i.ms ? i.ms + 'ms' : '—';
-    const up    = i.uptime_day != null ? i.uptime_day + '%' : '—';
-    const grade = i.https_grade || '—';
-    const gradeColor = grade.startsWith('A') ? '#3cb371' : grade === '—' ? '#888' : '#f0a000';
-    const engines = (i.engines_live && i.engines_live.length
-      ? i.engines_live : i.engines_meta || []).join(', ') || '—';
-    const btn = ok
-      ? `<button class="btn btn-secondary" style="font-size:11px;padding:3px 8px" onclick="select('${{i.url}}')">Seleziona</button>`
-      : '—';
-    tbody.innerHTML += `<tr>
-      <td style="font-family:monospace;font-size:11px">${{i.url}}</td>
-      <td><span style="color:${{color}}">${{stato}}</span></td>
-      <td>${{ms}}</td>
-      <td>${{i.country || '—'}}</td>
-      <td style="font-size:11px;color:#888">${{i.version || '—'}}</td>
-      <td>${{up}}</td>
-      <td style="color:${{gradeColor}}">${{grade}}</td>
-      <td style="font-size:11px;color:#888;max-width:200px;overflow:hidden;text-overflow:ellipsis">${{engines}}</td>
-      <td>${{btn}}</td>
-    </tr>`;
-  }});
-  document.getElementById('results-card').style.display = 'block';
-}}
-
-loadActive();
-</script>
-"""
-    return HTMLResponse(page("Istanze", content, "/instances/ui"))
-
-
 # ── Comandi ────────────────────────────────────────────────────────────────────
 
 @app.get("/commands", response_class=HTMLResponse)
@@ -309,17 +222,6 @@ def commands_page():
                 f'curl -N -X POST "{BASE}/ask/stream" -H "Content-Type: application/json" -d "{{\\"prompt\\":\\"chi e napoleone?\\"}}"',
                 f'curl -N -X POST "{BASE}/ask/stream" -H "Content-Type: application/json" -d "{{\\"prompt\\":\\"chi e napoleone?\\"}}"'),
         ],
-        "Istanze SearXNG": [
-            cmd_block("b3","Testa tutte le istanze",
-                f'Invoke-RestMethod "{BASE}/instances" | ConvertTo-Json -Depth 5',
-                f'curl "{BASE}/instances"'),
-            cmd_block("b4","Vedi istanza attiva",
-                f'Invoke-RestMethod "{BASE}/instances/active"',
-                f'curl "{BASE}/instances/active"'),
-            cmd_block("b5","Seleziona istanza (cambia URL)",
-                f'Invoke-RestMethod -Uri "{BASE}/instances/select?url=https://searx.be" -Method POST',
-                f'curl -X POST "{BASE}/instances/select?url=https://searx.be"'),
-        ],
         "Feedback": [
             cmd_block("b6","Risposta utile",
                 f'Invoke-RestMethod -Uri "{BASE}/feedback?positive=true" -Method POST',
@@ -332,7 +234,7 @@ def commands_page():
 
     sections_html = ""
     for section, items in blocks.items():
-        icon = {"Stato AI":"🟢","Chat":"💬","Istanze SearXNG":"🌐","Feedback":"👍"}.get(section,"")
+        icon = {"Stato AI":"🟢","Chat":"💬","Feedback":"👍"}.get(section,"")
         sections_html += f'<div style="margin-bottom:28px"><h2 style="color:#a0c4ff;font-size:15px;margin-bottom:12px;border-left:3px solid #7c9ef8;padding-left:10px">{icon} {section}</h2>{"".join(items)}</div>'
 
     content = f"""
@@ -379,17 +281,6 @@ function AskStream($d) {{
     curl -s -N -X POST "$BASE/ask/stream" -H "Content-Type: application/json" -d (@{{prompt=$d}}|ConvertTo-Json -Compress)
 }}
 function Stato {{ Invoke-RestMethod "$BASE/" | ConvertTo-Json }}
-function Istanze {{
-    Write-Host "Scansione istanze..." -ForegroundColor Yellow
-    $r = Invoke-RestMethod "$BASE/instances"
-    Write-Host "Online: $($r.online_with_results)/$($r.total_tested)" -ForegroundColor Green
-    $r.best | ForEach-Object {{ Write-Host "  $($_.url) $($_.ms)ms" -ForegroundColor Green }}
-    Write-Host "Attiva: $($r.active)" -ForegroundColor Cyan
-}}
-function Seleziona($url) {{
-    $r = Invoke-RestMethod -Uri "$BASE/instances/select?url=$url" -Method POST
-    Write-Host "Impostata: $($r.active)" -ForegroundColor Green
-}}
 function Feedback($p) {{
     Invoke-RestMethod -Uri "$BASE/feedback?positive=$p" -Method POST | Out-Null
     Write-Host "Feedback inviato!" -ForegroundColor Green
@@ -397,15 +288,13 @@ function Feedback($p) {{
 
 while ($true) {{
     Write-Host "`n===== GENERAI =====" -ForegroundColor Magenta
-    Write-Host "1. Domanda  2. Streaming  3. Stato  4. Istanze  5. Seleziona istanza  6/7. Feedback  0. Esci"
+    Write-Host "1. Domanda  2. Streaming  3. Stato  4/5. Feedback  0. Esci"
     switch (Read-Host "Scelta") {{
         "1" {{ Ask (Read-Host "Domanda") }}
         "2" {{ AskStream (Read-Host "Domanda") }}
         "3" {{ Stato }}
-        "4" {{ Istanze }}
-        "5" {{ Seleziona (Read-Host "URL istanza") }}
-        "6" {{ Feedback "true" }}
-        "7" {{ Feedback "false" }}
+        "4" {{ Feedback "true" }}
+        "5" {{ Feedback "false" }}
         "0" {{ exit }}
         default {{ Write-Host "Scelta non valida" -ForegroundColor Red }}
     }}
@@ -422,13 +311,11 @@ set BASE={BASE}
 :menu
 echo.
 echo ===== GENERAI =====
-echo 1.Domanda  2.Streaming  3.Stato  4.Istanze  5.Seleziona istanza  0.Esci
+echo 1.Domanda  2.Streaming  3.Stato  0.Esci
 set /p s=Scelta:
 if "%s%"=="1" (set /p d=Domanda: & curl -s -X POST "%BASE%/ask" -H "Content-Type: application/json" -d "{{\\"prompt\\":\\"%d%\\"}}" & echo.) & goto menu
 if "%s%"=="2" (set /p d=Domanda: & curl -N -X POST "%BASE%/ask/stream" -H "Content-Type: application/json" -d "{{\\"prompt\\":\\"%d%\\"}}" & echo.) & goto menu
 if "%s%"=="3" (curl -s "%BASE%/" & echo.) & goto menu
-if "%s%"=="4" (curl -s "%BASE%/instances" & echo.) & goto menu
-if "%s%"=="5" (set /p u=URL: & curl -s -X POST "%BASE%/instances/select?url=%u%" & echo.) & goto menu
 if "%s%"=="0" exit /b
 goto menu
 """
@@ -475,110 +362,6 @@ async def ask_stream(req: AskRequest):
 
     return StreamingResponse(generate(), media_type="text/event-stream")
 
-@app.get("/instances")
-async def instances_json():
-    import requests as req, time
-    from scraper import get_active_instance
-
-    def fetch_list():
-        try:
-            r = req.get("https://searx.space/data/instances.json", timeout=10)
-            data = r.json().get("instances", {})
-            result = {}
-            for url, info in data.items():
-                if info.get("network_type", "normal") != "normal":
-                    continue
-                result[url.rstrip("/")] = {
-                    "version":  info.get("version", "?"),
-                    "network":  info.get("network_type", "normal"),
-                    "uptime":   round(info.get("uptime", {}).get("uptimeDay", 0), 1) if isinstance(info.get("uptime"), dict) else None,
-                    "engines":  list(info.get("engines", {}).keys())[:8],
-                    "country":  info.get("network", {}).get("country", "?") if isinstance(info.get("network"), dict) else "?",
-                    "https":    info.get("tls", {}).get("grade", "?") if isinstance(info.get("tls"), dict) else "?",
-                }
-            return result
-        except Exception:
-            return {
-                "https://searx.be":          {"version":"?","network":"normal","uptime":None,"engines":[],"country":"BE","https":"?"},
-                "https://search.bus-hit.me": {"version":"?","network":"normal","uptime":None,"engines":[],"country":"?","https":"?"},
-                "https://searxng.site":       {"version":"?","network":"normal","uptime":None,"engines":[],"country":"?","https":"?"},
-            }
-
-    def check(url, meta):
-        try:
-            start = time.time()
-            r = req.get(f"{url}/search", params={"q":"test","format":"json"}, timeout=5,
-                        headers={"User-Agent":"GenerAI/1.0"})
-            ms = int((time.time()-start)*1000)
-            online = r.status_code == 200
-            try:
-                results_data = r.json()
-                has_results = len(results_data.get("results", [])) > 0
-                engines_used = list({e for res in results_data.get("results", []) for e in res.get("engines", [])})
-            except Exception:
-                has_results = False
-                engines_used = []
-            return {
-                "url":          url,
-                "online":       online,
-                "results":      has_results,
-                "ms":           ms,
-                "version":      meta.get("version", "?"),
-                "country":      meta.get("country", "?"),
-                "uptime_day":   meta.get("uptime"),
-                "https_grade":  meta.get("https", "?"),
-                "engines_live": engines_used[:6],
-                "engines_meta": meta.get("engines", []),
-                "network":      meta.get("network", "normal"),
-            }
-        except Exception as e:
-            return {
-                "url": url, "online": False, "results": False, "ms": None,
-                "version": meta.get("version","?"), "country": meta.get("country","?"),
-                "uptime_day": meta.get("uptime"), "https_grade": meta.get("https","?"),
-                "engines_live": [], "engines_meta": meta.get("engines",[]),
-                "network": meta.get("network","normal"), "error": str(e)[:60],
-            }
-
-    meta_map = await asyncio.to_thread(fetch_list)
-    checks = await asyncio.gather(*[asyncio.to_thread(check, u, m) for u, m in meta_map.items()])
-    srt = sorted(checks, key=lambda x: (not x["online"], not x.get("results"), x.get("ms") or 9999))
-    online = [c for c in srt if c["online"] and c.get("results")]
-    return {"active": get_active_instance(), "total_tested": len(srt),
-            "online_with_results": len(online), "best": online[:5], "all": srt}
-
-@app.get("/instances/active")
-def get_active():
-    from scraper import get_active_instance
-    return {"active": get_active_instance()}
-
-@app.get("/engines")
-def get_engines():
-    from scraper import get_engines
-    return {"engines": get_engines(), "available": ["brave","bing","google","duckduckgo","torch","yahoo","startpage","qwant"]}
-
-@app.post("/engines")
-def set_engines(engines: str):
-    from scraper import set_engines
-    set_engines(engines)
-    return {"ok": True, "engines": engines}
-
-@app.post("/instances/select")
-async def select_instance(url: str):
-    import requests as req
-    from scraper import set_active_instance
-    def test(u):
-        try:
-            r = req.get(f"{u}/search", params={"q":"test","format":"json"}, timeout=5,
-                        headers={"User-Agent":"GenerAI/1.0"})
-            return r.status_code == 200
-        except Exception:
-            return False
-    if not await asyncio.to_thread(test, url):
-        raise HTTPException(status_code=400, detail=f"Istanza non raggiungibile: {url}")
-    set_active_instance(url)
-    return {"ok": True, "active": url}
-
 @app.post("/feedback")
 def feedback(positive: bool):
     if brain:

requirements.txt

@@ -5,4 +5,4 @@ chromadb
 sentence-transformers
 trafilatura
 lxml_html_clean
-requests
+playwright

scraper.py

@@ -1,129 +1,118 @@
-import os
-import requests
+from urllib.parse import quote
 import trafilatura
+from playwright.sync_api import sync_playwright, TimeoutError as PlaywrightTimeoutError
 from errors import get_logger, GenerAIError, ErrorCode, fmt_exc
 
 log = get_logger("scraper")
 
-# Istanza SearXNG — configurabile via Secret SEARXNG_URL
-SEARXNG_URL = os.environ.get("SEARXNG_URL", "https://searx.be")
-_active_instance = SEARXNG_URL
+SEARCH_URL = "https://html.duckduckgo.com/html/"
+USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36"
+NAV_TIMEOUT_MS = 15000
 
-PUBLIC_INSTANCES = [
-    "https://searx.be",
-    "https://search.bus-hit.me",
-    "https://searxng.site",
-]
 
-# Motori attivi dentro SearXNG — brave, bing (Edge), torch, duckduckgo come fallback
-SEARXNG_ENGINES = os.environ.get("SEARXNG_ENGINES", "brave,bing,torch,duckduckgo")
+def _launch_browser(p):
+    return p.chromium.launch(
+        headless=True,
+        args=["--no-sandbox", "--disable-setuid-sandbox", "--disable-dev-shm-usage"],
+    )
 
-def get_active_instance() -> str:
-    return _active_instance
 
-def set_active_instance(url: str):
-    global _active_instance
-    _active_instance = url.rstrip("/")
-    log.info("Istanza SearXNG cambiata: %s", _active_instance)
+def _search_links(page, query: str, max_results: int) -> list[dict]:
+    """Cerca su DuckDuckGo (HTML, no JS) e restituisce [{title, url, snippet}]."""
+    url = f"{SEARCH_URL}?q={quote(query)}&kl=it-it"
+    page.goto(url, timeout=NAV_TIMEOUT_MS, wait_until="domcontentloaded")
 
-def get_engines() -> str:
-    return SEARXNG_ENGINES
-
-def set_engines(engines: str):
-    global SEARXNG_ENGINES
-    SEARXNG_ENGINES = engines
-    log.info("Motori SearXNG aggiornati: %s", SEARXNG_ENGINES)
-
-
-def _searxng_search(query: str, max_results: int) -> list[dict]:
-    """Cerca con SearXNG usando più motori. Prova più istanze se la prima fallisce."""
-    active = get_active_instance()
-    instances = [active] + [i for i in PUBLIC_INSTANCES if i != active]
-    engines = get_engines()
-
-    for instance in instances:
+    hits = []
+    rows = page.locator(".result").all()[: max_results * 2]
+    for row in rows:
         try:
-            resp = requests.get(
-                f"{instance}/search",
-                params={"q": query, "format": "json", "language": "it", "engines": engines},
-                timeout=8,
-                headers={"User-Agent": "GenerAI/1.0"},
-            )
-            if resp.status_code == 200:
-                data = resp.json()
-                hits = data.get("results", [])[:max_results * 2]
-                if hits:
-                    log.info("SearXNG (%s) [%s] → %d risultati", instance, engines, len(hits))
-                    return hits
-                # Nessun risultato con questi motori → riprova senza filtro motori
-                resp2 = requests.get(
-                    f"{instance}/search",
-                    params={"q": query, "format": "json", "language": "it"},
-                    timeout=8,
-                    headers={"User-Agent": "GenerAI/1.0"},
-                )
-                if resp2.status_code == 200:
-                    hits2 = resp2.json().get("results", [])[:max_results * 2]
-                    log.info("SearXNG (%s) [fallback tutti motori] → %d risultati", instance, len(hits2))
-                    return hits2
-        except Exception as e:
-            log.warning("SearXNG %s fallita: %s", instance, fmt_exc(e))
-    return []
+            link = row.locator("a.result__a").first
+            href = link.get_attribute("href")
+            title = link.inner_text().strip()
+            if not href or not href.startswith("http"):
+                continue
+            snippet_el = row.locator(".result__snippet").first
+            snippet = snippet_el.inner_text().strip() if snippet_el.count() else ""
+            hits.append({"title": title, "url": href, "snippet": snippet})
+        except Exception:
+            continue
+    return hits
+
+
+def _extract_text(page, url: str, snippet: str) -> str | None:
+    """Naviga alla pagina e ne estrae il testo pulito con trafilatura."""
+    try:
+        page.goto(url, timeout=NAV_TIMEOUT_MS, wait_until="domcontentloaded")
+        html = page.content()
+        text = trafilatura.extract(
+            html,
+            url=url,
+            include_links=False,
+            include_images=False,
+            include_tables=False,
+            no_fallback=False,
+        )
+    except PlaywrightTimeoutError:
+        log.warning("[%s] Timeout caricamento pagina: %s", ErrorCode.WEB_FETCH_FAILED.value, url)
+        text = None
+    except Exception as e:
+        log.warning("[%s] Fetch fallito per %s — %s", ErrorCode.WEB_FETCH_FAILED.value, url, fmt_exc(e))
+        text = None
 
+    # Fallback allo snippet del motore di ricerca se l'estrazione non basta
+    if not text or len(text) < 100:
+        text = snippet
 
-def search_and_extract(query: str, max_results: int = 3) -> list[dict]:
-    """Cerca con SearXNG ed estrae testo pulito. Returns [] on total failure."""
-    log.info("Ricerca web per: %r", query)
+    if not text or len(text) < 50:
+        return None
+    return text[:2000]
 
-    hits = _searxng_search(query, max_results)
 
-    if not hits:
-        log.warning("[%s] Nessun risultato SearXNG per: %r", ErrorCode.WEB_NO_RESULTS.value, query)
+def search_and_extract(query: str, max_results: int = 3) -> list[dict]:
+    """Cerca sul web tramite Chromium (Playwright) ed estrae testo pulito.
+
+    Ritorna [] se la ricerca fallisce completamente. Interfaccia invariata
+    rispetto alla precedente implementazione basata su SearXNG.
+    """
+    log.info("Ricerca web (Chromium) per: %r", query)
+
+    try:
+        with sync_playwright() as p:
+            browser = _launch_browser(p)
+            context = browser.new_context(user_agent=USER_AGENT)
+            page = context.new_page()
+            try:
+                hits = _search_links(page, query, max_results)
+            except PlaywrightTimeoutError:
+                log.warning("[%s] Timeout ricerca DuckDuckGo per: %r", ErrorCode.WEB_SEARCH_FAILED.value, query)
+                hits = []
+            except Exception as e:
+                log.warning("[%s] Ricerca DuckDuckGo fallita: %s", ErrorCode.WEB_SEARCH_FAILED.value, fmt_exc(e))
+                hits = []
+
+            if not hits:
+                browser.close()
+                log.warning("[%s] Nessun risultato per: %r", ErrorCode.WEB_NO_RESULTS.value, query)
+                return []
+
+            results = []
+            for hit in hits:
+                if len(results) >= max_results:
+                    break
+                log.debug("Fetching: %s", hit["url"])
+                text = _extract_text(page, hit["url"], hit["snippet"])
+                if not text:
+                    log.debug("Testo troppo corto per: %s", hit["url"])
+                    continue
+                log.info("Estratti %d chars da: %s", len(text), hit["url"])
+                results.append({"title": hit["title"], "url": hit["url"], "text": text})
+
+            browser.close()
+    except Exception as e:
+        err = GenerAIError(ErrorCode.WEB_SEARCH_FAILED, f"Browser Chromium non avviato: {fmt_exc(e)}", cause=e)
+        err.log(log)
         return []
 
-    results = []
-    for hit in hits:
-        if len(results) >= max_results:
-            break
-
-        # SearXNG include già uno snippet — usalo se il fetch fallisce
-        url     = hit.get("url", "")
-        title   = hit.get("title", "")
-        snippet = hit.get("content", "")
-
-        if not url:
-            continue
-
-        log.debug("Fetching: %s", url)
-        text = None
-        try:
-            downloaded = trafilatura.fetch_url(url, timeout=10)
-            if downloaded:
-                text = trafilatura.extract(
-                    downloaded,
-                    include_links=False,
-                    include_images=False,
-                    include_tables=False,
-                    no_fallback=False,
-                )
-        except Exception as e:
-            log.warning("[%s] fetch fallito per %s — %s", ErrorCode.WEB_FETCH_FAILED.value, url, fmt_exc(e))
-
-        # Fallback allo snippet SearXNG se trafilatura non estrae nulla
-        if not text or len(text) < 100:
-            text = snippet
-
-        if not text or len(text) < 50:
-            log.debug("Testo troppo corto per: %s", url)
-            continue
-
-        log.info("Estratti %d chars da: %s", len(text), url)
-        results.append({
-            "title": title,
-            "url": url,
-            "text": text[:2000],
-        })
-
     if not results:
         log.warning("[%s] Nessun testo utile estratto per: %r", ErrorCode.WEB_NO_RESULTS.value, query)