Update app.py

app.py

@@ -1,137 +1,121 @@
 import gradio as gr
 import pandas as pd
 import numpy as np
+import pickle
 import os
 import zipfile
-import glob
 from sentence_transformers import SentenceTransformer
 from sklearn.metrics.pairwise import cosine_similarity
 
-# --- 1. חילוץ תמונות חכם (לתוך תיקייה ייעודית) ---
-def setup_images():
-    if os.path.exists('images.zip'):
-        try:
-            with zipfile.ZipFile('images.zip', 'r') as zip_ref:
-                zip_ref.extractall('extracted_images')
-            print("✅ Images extracted successfully.")
-        except Exception as e:
-            print(f"❌ Zip Error: {e}")
+# --- 1. הכנה: פריקת תמונות ---
+if os.path.exists('images.zip'):
+    with zipfile.ZipFile('images.zip', 'r') as zip_ref:
+        zip_ref.extractall('extracted_images')
 
-setup_images()
-
-# --- 2. מנוע המלצות (סנכרון מלא 100%) ---
+# --- 2. טעינת נכסים (Assets) מקוריים מהקולאב ---
+# אנחנו משתמשים במודל רק כדי לקודד את החיפוש הנוכחי של המשתמש
 model = SentenceTransformer('all-MiniLM-L6-v2')
 
-def initialize_engine():
-    df = pd.read_csv('bitewise_clean_dataset.csv').fillna("N/A")
-    # יצירת הוקטורים מחדש מבטיחה שאינדקס X בטבלה הוא תמונה X
-    dish_texts = (df['dish_name'] + " " + df['food_vibe'] + " " + df['restaurant_name']).tolist()
-    dish_embeddings = model.encode(dish_texts, show_progress_bar=False)
-    return df, dish_embeddings
-
-main_df, dish_embeddings = initialize_engine()
-
-# פונקציית עזר למציאת נתיב התמונה המדויק
-def get_image_path(idx):
-    # מחפש את הקובץ בכל עומק התיקיות שחולצו
-    pattern = f"extracted_images/**/dish_{idx}_*.jpg"
-    files = glob.glob(pattern, recursive=True)
-    if files:
-        return f"file/{files[0]}"
-    return "https://via.placeholder.com/300?text=BiteWise+Dish"
-
-# --- 3. העיצוב היוקרתי (Premium Brown/Wood) ---
-custom_css = """
-@import url('https://fonts.googleapis.com/css2?family=Playfair+Display:wght@700&family=Inter:wght@400;600&display=swap');
-.gradio-container { background-color: #FAF9F6 !important; font-family: 'Inter', sans-serif !important; }
-.main-title { font-family: 'Playfair Display', serif; color: #3E2723; font-size: 3.5em; text-align: center; margin-bottom: 0px; }
-.sub-title { text-align: center; color: #7F4F24; letter-spacing: 2px; text-transform: uppercase; font-size: 0.8em; margin-bottom: 30px; }
+def load_data():
+    # קריאת הנתונים המדויקים שלכם
+    df = pd.read_csv('bitewise_clean_dataset.csv')
+    dish_embeddings = np.load('BiteWise_Dish_Embeddings.npy')
+    
+    with open('BiteWise_User_Embeddings.pkl', 'rb') as f:
+        user_embeddings = pickle.load(f)
+        if isinstance(user_embeddings, list):
+            user_embeddings = np.array(user_embeddings)
+            
+    # סנכרון אורכים למניעת תזוזות במידע
+    min_len = min(len(df), len(dish_embeddings), len(user_embeddings))
+    df = df.iloc[:min_len].copy()
+    dish_embeddings = dish_embeddings[:min_len]
+    user_embeddings = user_embeddings[:min_l]
+    
+    return df, dish_embeddings, user_embeddings
 
-.dish-card { 
-    background: white; border-radius: 15px; padding: 25px; margin-bottom: 25px; 
-    box-shadow: 0 10px 30px rgba(62, 39, 35, 0.08); border: 1px solid #EFEBE9;
-    display: flex; gap: 25px; border-left: 10px solid #3E2723;
-}
-.match-score { background: #3E2723; color: white; padding: 6px 16px; border-radius: 20px; font-weight: bold; font-size: 0.85em; }
-.info-tag { background: #F5F1EE; color: #5D4037; padding: 5px 12px; border-radius: 6px; font-weight: 700; font-size: 0.75em; text-transform: uppercase; border: 1px solid #D7CCC8; }
-.review-text { font-family: 'Playfair Display', serif; font-style: italic; color: #4E342E; font-size: 1.15em; margin: 15px 0; border-left: 2px solid #D7CCC8; padding-left: 15px; line-height: 1.5; }
-.twin-bio { background: #FFF9F0; padding: 15px; border-radius: 10px; font-size: 0.95em; color: #5D4037; border: 1px solid #FFE0B2; }
-"""
+# טעינה חד פעמית עם עליית האפליקציה
+try:
+    main_df, dish_embeddings, user_embeddings = load_data()
+except Exception as e:
+    print(f"Error loading files: {e}")
+    main_df, dish_embeddings, user_embeddings = None, None, None
 
-# --- 4. לוגיקה עסקית ---
+# --- 3. לוגיקת המלצה (העתק מהקולאב) ---
 def get_recommendations(query, origin, hobbies, style):
+    if main_df is None:
+        return "שגיאה: הקבצים המקוריים (CSV/NPY/PKL) חסרים בשרת."
+    
+    alpha = 0.7 # המשקולת המקורית שלכם
+    
+    # וקטוריזציה של הקלט הנוכחי
     q_vec = model.encode([str(query)])
-    d_sim = cosine_similarity(q_vec, dish_embeddings).flatten()
+    u_input_str = f"Origin: {origin}, Hobbies: {hobbies}, Style: {style}"
+    u_input_vec = model.encode([u_input_str])
+    
+    # חישוב דמיון מול הוקטורים המוכנים שלכם
+    dish_sim = cosine_similarity(q_vec, dish_embeddings).flatten()
+    user_sim = cosine_similarity(u_input_vec, user_embeddings).flatten()
     
-    res = main_df.copy()
-    res['score'] = d_sim
-    # מניעת כפילויות: מסנן אם אותה מנה מופיעה פעמיים מאותה מסעדה
-    top_results = res.sort_values('score', ascending=False).drop_duplicates(subset=['dish_name', 'restaurant_name']).head(3)
+    # ציון משולב
+    scores = (dish_sim * alpha) + (user_sim * (1 - alpha))
+    
+    temp_df = main_df.copy()
+    temp_df['final_score'] = scores
+    
+    # מניעת כפילויות: לוקחים את המנה הכי טובה מכל מסעדה
+    top_3 = temp_df.sort_values('final_score', ascending=False).drop_duplicates('restaurant_name').head(3)
     
     html = ""
-    for idx, row in top_results.iterrows():
-        pct = f"{min(98.8, 88 + (row['score'] * 20)):.1f}%"
-        img_url = get_image_path(idx)
+    for idx, row in top_3.iterrows():
+        # אחוז התאמה לפי הלוגיקה שלכם
+        pct = f"{min(98.8, 88 + (row['final_score'] * 20)):.1f}%"
+        
+        # חיפוש תמונה בתוך התיקייה שפרקנו
+        img_filename = f"dish_{idx}_0.jpg"
+        img_path = os.path.join('extracted_images', img_filename)
+        img_url = f"file/{img_path}" if os.path.exists(img_path) else "https://via.placeholder.com/300?text=BiteWise"
         
         html += f"""
-        <div class="dish-card">
-            <img src="{img_url}" style="width: 280px; height: 280px; object-fit: cover; border-radius: 12px;">
+        <div style="border: 2px solid #3E2723; border-radius: 15px; padding: 20px; margin-bottom: 20px; display: flex; gap: 20px; background: white;">
+            <img src="{img_url}" style="width: 250px; height: 250px; object-fit: cover; border-radius: 10px;">
             <div style="flex: 1;">
-                <div style="display: flex; justify-content: space-between; align-items: center;">
-                    <h2 style="margin: 0; font-family: 'Playfair Display', serif; color: #3E2723; font-size: 2em;">{row['dish_name']}</h2>
-                    <span class="match-score">{pct} MATCH</span>
-                </div>
-                <p style="color: #7F4F24; font-weight: 600; margin: 5px 0;">📍 {row['restaurant_name']}</p>
-                <div style="margin: 15px 0; display: flex; gap: 10px;">
-                    <span class="info-tag">✧ {row['food_vibe']}</span>
-                    <span class="info-tag">✧ {row['ideal_context']}</span>
-                    <span class="info-tag">✧ RATING: {row['rating']}</span>
-                </div>
-                <div class="review-text">"{row['taste_review']}"</div>
-                <div class="twin-bio">
-                    <strong>Twin Bio:</strong> Your culinary twin from <b>{origin}</b> who shares your interest in <b>{hobbies}</b> recommends this <b>{style}</b> experience.
+                <h2 style="margin: 0; color: #3E2723;">{row['dish_name']} <small>@{row['restaurant_name']}</small></h2>
+                <p style="font-weight: bold; color: #7F4F24; font-size: 1.1em;">{pct} Match</p>
+                <div style="margin: 10px 0;">
+                    <span style="background: #F5F1EE; padding: 5px 10px; border-radius: 5px; font-weight: bold;">🍽️ {row['food_vibe']}</span>
+                    <span style="background: #F5F1EE; padding: 5px 10px; border-radius: 5px; font-weight: bold; margin-left: 5px;">👥 {row['ideal_context']}</span>
                 </div>
+                <p style="font-style: italic; color: #444;">"{row['taste_review']}"</p>
+                <hr style="border: 0; border-top: 1px solid #eee; margin: 10px 0;">
+                <p style="margin: 0; font-size: 0.9em;"><b>Rating:</b> {row['rating']}/5 | <b>Twin:</b> Based on {style} DNA</p>
             </div>
         </div>
         """
     return html
 
-def handle_share(dish, rest, review):
-    if not dish or not rest or not review:
-        return "### ⚠️ Please fill in all fields before uploading."
-    return f"### ✅ Success!\n**{dish}** from **{rest}** has been shared with the community. Your review helps others find their Culinary Twin!"
+# --- 4. הממשק היוקרתי (CSS) ---
+custom_css = """
+.gradio-container { background-color: #FAF9F6 !important; }
+.main-title { color: #3E2723; font-family: 'Serif'; text-align: center; font-size: 3em; }
+"""
 
-# --- 5. ממשק המשתמש ---
 with gr.Blocks(css=custom_css) as demo:
-    gr.HTML("<div class='main-title'>BiteWise</div>")
-    gr.HTML("<div class='sub-title'>Curated Culinary Twin Experiences</div>")
+    gr.HTML("<h1 class='main-title'>BiteWise</h1>")
     
     with gr.Tabs() as tabs:
         with gr.Tab("01. YOUR DNA", id=0):
             with gr.Row():
-                u_n = gr.Textbox(label="Full Name", placeholder="What's your name?")
-                u_o = gr.Dropdown(list(main_df['user_origin'].unique()), label="Origin", value="Tel Aviv")
-            with gr.Row():
-                u_h = gr.Dropdown(list(main_df['user_hobbies'].unique()), label="Hobbies", value="Photography")
-                u_s = gr.Dropdown(list(main_df['user_fashion_style'].unique()), label="Style", value="Modern/Minimalist")
-            btn1 = gr.Button("SYNC MY DNA", variant="primary")
+                u_o = gr.Dropdown(["Tel Aviv", "London", "New York", "Paris", "Berlin", "Tokyo"], label="Origin", value="Tel Aviv")
+                u_h = gr.Dropdown(["Photography", "Cooking", "Fashion", "Traveling", "Gaming", "Yoga"], label="Hobbies", value="Photography")
+                u_s = gr.Dropdown(["Modern/Minimalist", "Classic/Elegant", "Street/Urban", "Nature/Boho"], label="Style", value="Modern/Minimalist")
+            btn1 = gr.Button("SAVE DNA", variant="primary")
             btn1.click(lambda: gr.update(selected=1), None, tabs)
 
         with gr.Tab("02. DISCOVERY", id=1):
-            with gr.Row():
-                q_in = gr.Textbox(label="What are you craving?", placeholder="e.g. Sushi, Pasta...", scale=4)
-                btn2 = gr.Button("FIND", variant="primary", scale=1)
-            out_html = gr.HTML()
-            btn2.click(get_recommendations, [q_in, u_o, u_h, u_s], out_html)
-
-        with gr.Tab("03. SHARE", id=2):
-            gr.Markdown("### 📸 Share your discovery with the community")
-            with gr.Column():
-                s_dish = gr.Textbox(label="Dish Name")
-                s_rest = gr.Textbox(label="Restaurant")
-                s_rev = gr.Textbox(label="Your Review", lines=3)
-                btn_up = gr.Button("UPLOAD TO COMMUNITY")
-                s_msg = gr.Markdown()
-                btn_up.click(handle_share, [s_dish, s_rest, s_rev], s_msg)
+            q_in = gr.Textbox(label="What are you craving?", placeholder="e.g. Pasta")
+            btn2 = gr.Button("FIND", variant="primary")
+            out = gr.HTML()
+            btn2.click(get_recommendations, [q_in, u_o, u_h, u_s], out)
 
 demo.launch(allowed_paths=["."])