PhilipL commited on
Commit
40e5154
·
verified ·
1 Parent(s): 4aba2f8

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +53 -70
app.py CHANGED
@@ -74,69 +74,69 @@ def init_static_data():
74
  # ==========================================
75
  def sync_google_maps(url):
76
  global global_df, global_retriever, rag_initialized
77
-
78
- # 清理網址
79
  clean_url = url.strip() if url else ""
80
-
81
  if not clean_url or "http" not in clean_url.lower():
82
- yield "❌ 請輸入包含 http 的有效網址。"
83
  return
84
 
85
  try:
86
- yield "🚀 [1/4] 正在啟動瀏覽器..."
87
  with sync_playwright() as p:
88
  browser = p.chromium.launch(headless=True)
89
- context = browser.new_context(
90
- viewport={'width': 1280, 'height': 800},
91
- 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"
92
- )
93
- page = context.new_page()
94
-
95
- yield "🌐 [2/4] 正在連線至地圖清單..."
96
- # 使用 domcontentloaded 較快,後續配合 time.sleep 穩定性最高
97
  page.goto(clean_url, wait_until="domcontentloaded", timeout=60000)
 
98
 
99
- yield "⏳ [3/4] 正在等待網頁內容加載 (約 10 秒)..."
100
- time.sleep(10) # ✅ 現在已經 import time 了,不會再報錯
101
 
102
- yield "📄 [4/4] 正在解析地點..."
103
-
104
- # 抓取店名與地址
105
- titles = page.locator('div.fontHeadlineSmall').all_inner_texts()
106
- details = page.locator('div.fontBodyMedium').all_inner_texts()
107
 
108
  restaurant_list = []
109
- for i, name in enumerate(titles):
110
- name = name.strip()
111
- if name:
112
- addr = details[i].strip() if i < len(details) else "詳情請見地圖"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
113
  restaurant_list.append({
114
  "Name": name,
115
  "Address": addr,
116
- "RAG_Content": f"推薦餐廳:{name},地點資訊:{addr}",
117
- # ✅ 使用 urllib.parse 確保中文店名不會導致連結斷掉
118
  "URL": f"https://www.google.com/search?q={urllib.parse.quote(name)}"
119
  })
120
-
 
 
121
  browser.close()
122
 
123
  if not restaurant_list:
124
- yield "⚠️ 同步完成但找不到餐廳。請確認連結格式正確且清單已設為「共用」。"
125
  return
126
 
127
- yield f"📥 成功提取 {len(restaurant_list)} 個地點,正在更新向量資料庫..."
128
-
129
  global_df = pd.DataFrame(restaurant_list)
130
  embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
131
- texts = global_df['RAG_Content'].tolist()
132
- metadatas = global_df.to_dict('records')
133
-
134
- vectorstore = FAISS.from_texts(texts, embeddings, metadatas=metadatas)
135
  global_retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
136
-
137
  rag_initialized = True
138
- yield f"✅ ��步完成!目前資料庫已有 {len(restaurant_list)} 間餐廳。"
139
-
140
  except Exception as e:
141
  yield f"❌ 系統錯誤:{str(e)}"
142
 
@@ -193,48 +193,31 @@ def generate_image_huggingface(prompt):
193
  except: return None
194
 
195
  def mood_agent_logic(score_input, food_input, diary_input, debug_mode):
196
- if not rag_initialized or global_df is None:
197
- yield "⚠️ 請先在上方貼上 Google Maps 連結並按下『同步我的清單』!", None, ""
198
  return
199
 
200
- # 取得資料
201
  restaurant, is_match, rec_categories, mood_reason, all_names = get_restaurant_data(score_input, food_input)
202
 
203
- name = restaurant['Name']; address = restaurant['Address']; url = restaurant.get('URL', '#')
204
- rag_info = str(restaurant.get('RAG_Content', ''))
205
-
206
- # RAG 檢索增強
207
- if global_retriever:
208
- try:
209
- docs = global_retriever.invoke(name)
210
- if docs: rag_info = "\n".join([d.page_content for d in docs])
211
- except: pass
212
-
213
- # 構造提示詞:如果沒匹配到,要告訴 AI 讓它解釋原因
214
- match_status_msg = ""
215
- if not is_match and food_input != "隨便":
216
- match_status_msg = f"\n(注意:使用者的資料庫中找不到符合 '{food_input}' 的餐廳。請在回應中幽默地提到:雖然目前的清單裡只有這幾間店:[{all_names}],但你還是推薦了這間。)"
217
-
218
- ai_text = generate_content_with_groq(
219
- name,
220
- rag_info + match_status_msg,
221
- diary_input,
222
- score_input,
223
- mood_reason,
224
- debug_mode
225
- )
226
-
227
- # 建立最終回傳訊息
228
  prefix = ""
229
  if not is_match and food_input != "隨便":
230
- prefix = f"> 💡 **小提醒**:在您的清單中沒找到『{food_input}』類型的餐廳,目前您的地圖清單共有:**{all_names}**。所以我先從現有的為您挑選一間!\n\n"
 
 
 
 
 
 
 
 
231
 
232
  final_response = f"{prefix}### 🍽️ 推薦:{name}\n\n{ai_text}"
233
- 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">🗺️ 點我查詢或導航</a></div>'
234
 
235
- yield final_response, None, map_html
236
- image_output = generate_image_huggingface(f"Gourmet food: {name}")
237
- yield final_response, image_output, map_html
238
 
239
  # ==========================================
240
  # 3. 橋接邏輯 (保留原本情緒辨識部分)
 
74
  # ==========================================
75
  def sync_google_maps(url):
76
  global global_df, global_retriever, rag_initialized
 
 
77
  clean_url = url.strip() if url else ""
 
78
  if not clean_url or "http" not in clean_url.lower():
79
+ yield "❌ 請輸入有效的網址。"
80
  return
81
 
82
  try:
83
+ yield "🚀 [1/4] 啟動瀏覽器並準備抓取實景圖..."
84
  with sync_playwright() as p:
85
  browser = p.chromium.launch(headless=True)
86
+ page = browser.new_page()
 
 
 
 
 
 
 
87
  page.goto(clean_url, wait_until="domcontentloaded", timeout=60000)
88
+ time.sleep(10) # 等待圖片加載
89
 
90
+ yield "📄 [2/4] 正在解析餐廳資訊與圖片..."
 
91
 
92
+ # 定位所有的餐廳卡片
93
+ items = page.locator('div[role="article"]').all()
 
 
 
94
 
95
  restaurant_list = []
96
+ for item in items:
97
+ try:
98
+ # 抓取店名
99
+ name_el = item.locator('div.fontHeadlineSmall')
100
+ if name_el.count() == 0: continue
101
+ name = name_el.inner_text().strip()
102
+
103
+ # 抓取地址/描述
104
+ detail_el = item.locator('div.fontBodyMedium').first
105
+ addr = detail_el.inner_text().strip() if detail_el.count() > 0 else "位置詳見地圖"
106
+
107
+ # ✅ 抓取圖片網址 (Google Maps 的縮圖通常在 img 標籤中)
108
+ img_el = item.locator('img').first
109
+ img_url = None
110
+ if img_el.count() > 0:
111
+ img_url = img_el.get_attribute('src')
112
+
113
+ # 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
114
+ if not img_url or img_url.startswith('data:'):
115
+ img_url = "https://images.unsplash.com/photo-1517248135467-4c7edcad34c4?q=80&w=400" # 預設餐廳美圖
116
+
117
  restaurant_list.append({
118
  "Name": name,
119
  "Address": addr,
120
+ "Image_URL": img_url, # ✅ 新增欄位
121
+ "RAG_Content": f"餐廳:{name},資訊:{addr}",
122
  "URL": f"https://www.google.com/search?q={urllib.parse.quote(name)}"
123
  })
124
+ except:
125
+ continue
126
+
127
  browser.close()
128
 
129
  if not restaurant_list:
130
+ yield "⚠️ 找不到餐廳資訊,請確認連結格式。"
131
  return
132
 
133
+ # 更新向量資料庫
 
134
  global_df = pd.DataFrame(restaurant_list)
135
  embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
136
+ vectorstore = FAISS.from_texts(global_df['RAG_Content'].tolist(), embeddings, metadatas=global_df.to_dict('records'))
 
 
 
137
  global_retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
 
138
  rag_initialized = True
139
+ yield f"✅ 同步成功!已載入 {len(restaurant_list)} 間包含實景圖的餐廳。"
 
140
  except Exception as e:
141
  yield f"❌ 系統錯誤:{str(e)}"
142
 
 
193
  except: return None
194
 
195
  def mood_agent_logic(score_input, food_input, diary_input, debug_mode):
196
+ if not rag_initialized:
197
+ yield "⚠️ 請先同步地圖清單!", None, ""
198
  return
199
 
200
+ # 取得推薦資料
201
  restaurant, is_match, rec_categories, mood_reason, all_names = get_restaurant_data(score_input, food_input)
202
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
203
  prefix = ""
204
  if not is_match and food_input != "隨便":
205
+ 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"
206
+
207
+ name = restaurant['Name']
208
+ url = restaurant.get('URL', '#')
209
+ real_img_url = restaurant.get('Image_URL') # ✅ 取得抓到的真實圖片
210
+ rag_info = str(restaurant.get('RAG_Content', ''))
211
+
212
+ # 產生 AI 回應
213
+ ai_text = generate_content_with_groq(name, rag_info, diary_input, score_input, mood_reason)
214
 
215
  final_response = f"{prefix}### 🍽️ 推薦:{name}\n\n{ai_text}"
216
+ 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>'
217
 
218
+ # ✅ 直接回傳圖片網址給 Gradio 的 image_output
219
+ # 注意:Gradio 的 gr.Image 可以直接接收 URL 字串
220
+ yield final_response, real_img_url, map_html
221
 
222
  # ==========================================
223
  # 3. 橋接邏輯 (保留原本情緒辨識部分)