Update app.py

app.py

@@ -9,10 +9,7 @@ import subprocess
 os.environ["PLAYWRIGHT_BROWSERS_PATH"] = "/home/user/.cache/ms-playwright"
 
 try:
-    # 如果目錄不存在，才執行下載，節省啟動時間
     if not os.path.exists(os.environ["PLAYWRIGHT_BROWSERS_PATH"]):
-        print("正在為 Debian Trixie 環境安裝 Chromium...")
-        # 僅安裝 chromium 即可，依賴項由 packages.txt 處理
         subprocess.run(["python3", "-m", "playwright", "install", "chromium"], check=True)
 except Exception as e:
     print(f"Playwright 初始化提示: {e}")
@@ -83,57 +80,64 @@ def sync_google_maps(url):
         yield "🚀 [1/4] 啟動瀏覽器並準備抓取實景圖..."
         with sync_playwright() as p:
             browser = p.chromium.launch(headless=True)
-            page = browser.new_page()
+            context = browser.new_context(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")
+            page = context.new_page()
+            yield "🌐 [2/4] 連線中..."
             page.goto(clean_url, wait_until="domcontentloaded", timeout=60000)
-            time.sleep(10) # 等待圖片加載
+            yield "⏳ [3/4] 等待網頁內容渲染 (10秒)..."
+            time.sleep(10) 
             
-            yield "📄 [2/4] 正在解析餐廳資訊與圖片..."
-            
-            # 定位所有的餐廳卡片
-            items = page.locator('div[role="article"]').all()
+            yield "📄 [4/4] 正在解析餐廳資訊與圖片網址..."
+            # 【關鍵修改】改用名稱作為基準，再去找對應的圖片，避免因為 role="article" 抓不到
+            name_elements = page.locator('div.fontHeadlineSmall').all()
             
             restaurant_list = []
-            for item in items:
+            for el in name_elements:
                 try:
-                    # 抓取店名
-                    name_el = item.locator('div.fontHeadlineSmall')
-                    if name_el.count() == 0: continue
-                    name = name_el.inner_text().strip()
+                    name = el.inner_text().strip()
+                    if not name: continue
                     
-                    # 抓取地址/描述
-                    detail_el = item.locator('div.fontBodyMedium').first
-                    addr = detail_el.inner_text().strip() if detail_el.count() > 0 else "位置詳見地圖"
+                    # 尋找同一層級或父層級中的圖片與地址
+                    # Google Maps 的結構通常是名稱在一個 div，圖片在旁邊的 div
+                    container = page.locator(f"//div[contains(., '{name}')]/ancestor::div[@role='article']").first
                     
-                    # ✅ 抓取圖片網址 (Google Maps 的縮圖通常在 img 標籤中)
-                    img_el = item.locator('img').first
-                    img_url = None
-                    if img_el.count() > 0:
-                        img_url = img_el.get_attribute('src')
+                    # 抓取地址
+                    addr = "位置詳見地圖"
+                    addr_el = container.locator('div.fontBodyMedium').first
+                    if addr_el.count() > 0:
+                        addr = addr_el.inner_text().strip()
                     
-                    # 如果圖片是 Data URL (base64) 或太小，通常代表沒載入成功，可以設個預設圖
-                    if not img_url or img_url.startswith('data:'):
-                        img_url = "https://images.unsplash.com/photo-1517248135467-4c7edcad34c4?q=80&w=400" # 預設餐廳美圖
+                    # 【新增】抓取實景圖 URL
+                    img_url = "https://images.unsplash.com/photo-1517248135467-4c7edcad34c4?q=80&w=400" # 預設圖
+                    img_el = container.locator('img').first
+                    if img_el.count() > 0:
+                        src = img_el.get_attribute('src')
+                        if src and not src.startswith('data:'):
+                            # 將縮圖網址改為較大尺寸
+                            img_url = src.replace('=w80-h80', '=w400-h400').split('=')[0] + '=w400-h400'
 
                     restaurant_list.append({
                         "Name": name,
                         "Address": addr,
-                        "Image_URL": img_url, # ✅ 新增欄位
-                        "RAG_Content": f"餐廳：{name}，資訊：{addr}",
+                        "Image_URL": img_url, # 【新增】存入圖片網址
+                        "RAG_Content": f"餐廳名稱：{name}，地址：{addr}",
                         "URL": f"https://www.google.com/search?q={urllib.parse.quote(name)}"
                     })
-                except:
-                    continue
-                    
+                except: continue
+                
             browser.close()
 
             if not restaurant_list:
-                yield "⚠️ 找不到餐廳資訊，請確認連結格式。"
+                yield "⚠️ 找不到餐廳資訊。請確認：1.連結是『分享清單』產生的 2.清單已設為公開。"
                 return
 
             # 更新向量資料庫
             global_df = pd.DataFrame(restaurant_list)
             embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
-            vectorstore = FAISS.from_texts(global_df['RAG_Content'].tolist(), embeddings, metadatas=global_df.to_dict('records'))
+            texts = global_df['RAG_Content'].tolist()
+            metadatas = global_df.to_dict('records')
+            
+            vectorstore = FAISS.from_texts(texts, embeddings, metadatas=metadatas)
             global_retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
             rag_initialized = True
             yield f"✅ 同步成功！已載入 {len(restaurant_list)} 間包含實景圖的餐廳。"
@@ -147,6 +151,9 @@ def get_restaurant_data(mood_score_str, food_choice):
     init_static_data()
     if global_df is None or global_df.empty: 
         return None, True, "資料庫未載入", "尚未同步地圖資料", ""
+
+    # 【新增】列出目前資料庫中所有餐廳名稱
+    all_names = "、".join(global_df['Name'].tolist())
     
     # 取得目前資料庫中所有餐廳的名稱清單，方便回報給使用者
     all_restaurant_names = "、".join(global_df['Name'].tolist())
@@ -200,14 +207,18 @@ def mood_agent_logic(score_input, food_input, diary_input, debug_mode):
     # 取得推薦資料
     restaurant, is_match, rec_categories, mood_reason, all_names = get_restaurant_data(score_input, food_input)
     
+    # 【新增】處理「找不到符合要求」時的告知訊息
     prefix = ""
+    status_hint = ""
     if not is_match and food_input != "隨便":
-        prefix = f"> 💡 **溫馨提示**：名單中沒找到『{food_input}』，目前清單有：**{all_names}**。為您挑選了這一間：\n\n"
+        prefix = f"> 💡 **小提醒**：在您的清單中沒找到『{food_input}』相關的店，目前名單有：**{all_names}**。所以我先推薦這間給您：\n\n"
+        status_hint = f"\n(注意：使用者的清單中無 '{food_input}'，請幽默解釋為何這間也值得一試。)"
 
     name = restaurant['Name']
     url = restaurant.get('URL', '#')
-    real_img_url = restaurant.get('Image_URL') # ✅ 取得抓到的真實圖片
-    rag_info = str(restaurant.get('RAG_Content', ''))
+    # 【修改】使用抓取到的真實圖片 URL
+    real_img_url = restaurant.get('Image_URL') 
+    rag_info = str(restaurant.get('RAG_Content', '')) + status_hint
     
     # 產生 AI 回應
     ai_text = generate_content_with_groq(name, rag_info, diary_input, score_input, mood_reason)
@@ -215,8 +226,7 @@ def mood_agent_logic(score_input, food_input, diary_input, debug_mode):
     final_response = f"{prefix}### 🍽️ 推薦：{name}\n\n{ai_text}"
     map_html = f'<div style="text-align:center"><a href="{url}" target="_blank" style="background:#4CAF50;color:white;padding:8px 16px;border-radius:20px;text-decoration:none">🗺️ 開啟 Google Map 導航</a></div>'
     
-    # ✅ 直接回傳圖片網址給 Gradio 的 image_output
-    # 注意：Gradio 的 gr.Image 可以直接接收 URL 字串
+    # 【修改】直接回傳實景圖網址給 Gradio
     yield final_response, real_img_url, map_html
 
 # ==========================================