Update app.py

app.py

@@ -1,64 +1,105 @@
+import warnings
+warnings.filterwarnings("ignore")
 
 import gradio as gr
 import pandas as pd
 import numpy as np
 import os
-import random
-from sklearn.feature_extraction.text import TfidfVectorizer
-from sklearn.pipeline import Pipeline
+import json
+from sentence_transformers import SentenceTransformer
 from sklearn.linear_model import LogisticRegression
+from datasets import load_dataset
 from huggingface_hub import InferenceClient
 
-# --- LIBRARIES & LOGIC ---
-print("🚀 Generating Synthetic Data...")
-proteins = ["Chicken", "Beef", "Tofu", "Salmon", "Shrimp", "Chickpeas", "Lentils", "Turkey", "Duck", "Egg"]
-carbs = ["Rice", "Pasta", "Quinoa", "Potatoes", "Bread", "Noodles", "Tortilla", "Bun", "Couscous"]
-veggies = ["Broccoli", "Spinach", "Carrots", "Tomatoes", "Mushrooms", "Onions", "Peppers", "Kale", "Avocado", "Corn"]
-sauces = ["Tomato Sauce", "Cream", "Soy Sauce", "Pesto", "BBQ Sauce", "Tahini", "Olive Oil", "Balsamic", "Curry", "Mayo"]
-spices = ["Garlic", "Basil", "Chili", "Curry Powder", "Cumin", "Oregano", "Ginger", "Thyme", "Paprika"]
-adjectives = ["Spicy", "Creamy", "Roasted", "Grilled", "Fresh", "Zesty", "Hearty", "Crispy", "Savory", "Sweet"]
-
-def generate_recipe():
-    p = random.choice(proteins); c = random.choice(carbs); v1 = random.choice(veggies); s = random.choice(sauces); sp = random.choice(spices); adj = random.choice(adjectives)
-    title = f"{adj} {p} with {c}"
-    ingredients = f"{p}, {c}, {v1}, {s}, {sp}"
-    vibe = "Quick Lunch"
-    if "Salmon" in p or "Shrimp" in p or "Duck" in p: vibe = "Fancy Dinner"
-    elif "Chocolate" in ingredients or "Risotto" in title: vibe = "Romantic"
-    elif "Tofu" in p or "Kale" in v1 or "Quinoa" in c: vibe = "Healthy Boost"
-    elif "Tortilla" in c or "BBQ" in s or "Crispy" in title: vibe = "Party Snack"
-    elif "Potatoes" in c or "Cream" in s or "Pasta" in c: vibe = "Comfort Food"
-    return {"title": title, "ingredients": ingredients, "Vibe_AI": vibe}
-
-data = [generate_recipe() for _ in range(5000)]
-df = pd.DataFrame(data)
-X = df['ingredients']; y = df['Vibe_AI']
-pipeline = Pipeline([('tfidf', TfidfVectorizer(stop_words='english')),('clf', LogisticRegression(max_iter=1000))])
-pipeline.fit(X, y)
-
+# ============================================
+# 1. Load Dataset from HuggingFace
+# ============================================
+print("Loading dataset from HuggingFace...")
+ds = load_dataset("adigabay2003/smartchef-recipes", split="train")
+df = ds.to_pandas()
+df = df.dropna(subset=["text"])
+df = df[df["text"].str.strip() != ""]
+df = df.reset_index(drop=True)
+print(f"Loaded {len(df)} recipes")
+
+# ============================================
+# 2. Load Optimal Model from JSON
+# ============================================
+with open("optimal_model.json", "r") as f:
+    optimal = json.load(f)
+
+MODEL_NAME = optimal["model_name"]
+print(f"Using embedding model: {MODEL_NAME}")
+
+# ============================================
+# 3. Load Precomputed Embeddings
+# ============================================
+print("Loading precomputed embeddings...")
+X = np.load("best_embeddings.npy")
+y = df["vibe"].tolist()
+print(f"Embeddings shape: {X.shape}")
+
+# ============================================
+# 4. Train Classifier on Embeddings
+# ============================================
+print("Training classifier...")
+embedder = SentenceTransformer(MODEL_NAME)
+clf = LogisticRegression(max_iter=1000)
+clf.fit(X, y)
+print("Classifier ready!")
+
+# ============================================
+# 5. Prediction Function
+# ============================================
 def get_recipe_vibe(title, ingredients):
-    full_text = f"{title} {ingredients}"
-    return pipeline.predict([full_text])[0]
-
+    text = f"{title}. Ingredients: {ingredients}"
+    embedding = embedder.encode([text])
+    vibe = clf.predict(embedding)[0]
+    return vibe
+
+# ============================================
+# 6. Image Generation
+# ============================================
 def generate_food_image(title, vibe):
     token = os.getenv("HF_TOKEN")
-    if not token: return None
-    client = InferenceClient(model="stabilityai/stable-diffusion-xl-base-1.0", token=token)
-    prompt = f"Professional food photography of {title}, {vibe} style, 4k, highly detailed, appetizing, dramatic lighting, vibrant colors"
-    try: return client.text_to_image(prompt)
-    except: return None
-
+    if not token:
+        return None
+    client = InferenceClient(
+        model="stabilityai/stable-diffusion-xl-base-1.0",
+        token=token
+    )
+    prompt = (
+        f"Professional food photography of {title}, "
+        f"{vibe} style, 4k, highly detailed, "
+        f"appetizing, dramatic lighting, vibrant colors"
+    )
+    try:
+        return client.text_to_image(prompt)
+    except:
+        return None
+
+# ============================================
+# 7. Main App Function
+# ============================================
 def smart_chef_app(title, ingredients):
     vibe = get_recipe_vibe(title, ingredients)
-    messages = {"Romantic": "🌹 Love is in the air!", "Quick Lunch": "⚡ Fast & Delicious!", "Comfort Food": "🧸 Warm & Cozy.", "Party Snack": "🎉 Party Time!", "Healthy Boost": "🥗 Feel Good Food.", "Fancy Dinner": "🍷 Chef's Kiss!"}
+    messages = {
+        "Romantic":     "🌹 Love is in the air!",
+        "Quick Lunch":  "⚡ Fast & Delicious!",
+        "Comfort Food": "🧸 Warm & Cozy.",
+        "Party Snack":  "🎉 Party Time!",
+        "Healthy Boost":"🥗 Feel Good Food.",
+        "Fancy Dinner": "🍷 Chef's Kiss!"
+    }
     msg = messages.get(vibe, "Yum!")
     img = generate_food_image(title, vibe)
     return vibe, msg, img
 
-# --- THE NEW ULTRA-PREMIUM UI CSS ---
+# ============================================
+# 8. CSS & Theme
+# ============================================
 ultra_css = """
 @import url('https://fonts.googleapis.com/css2?family=Montserrat:wght@300;500;800&display=swap');
-
 .gradio-container {
     background: linear-gradient(-45deg, #0f0c29, #302b63, #24243e, #4a1c1c);
     background-size: 400% 400%;
@@ -126,6 +167,9 @@ theme = gr.themes.Soft(primary_hue="orange", neutral_hue="slate").set(
     block_border_width="0px"
 )
 
+# ============================================
+# 9. Gradio UI
+# ============================================
 with gr.Blocks(css=ultra_css, theme=theme) as demo:
     gr.Markdown("# ✨ SMARTCHEF AI ✨")
     gr.Markdown("### Experience the future of culinary magic.")
@@ -133,23 +177,39 @@ with gr.Blocks(css=ultra_css, theme=theme) as demo:
     with gr.Row():
         with gr.Column(elem_classes="group-container"):
             gr.Markdown("#### 📝 CREATE YOUR DISH")
-            t_in = gr.Textbox(label="Dish Name", placeholder="e.g., Mystical Truffle Risotto")
-            i_in = gr.Textbox(label="Ingredients", placeholder="e.g., Arborio rice, black truffle dust, parmesan magic...", lines=4)
+            t_in = gr.Textbox(
+                label="Dish Name",
+                placeholder="e.g., Mystical Truffle Risotto"
+            )
+            i_in = gr.Textbox(
+                label="Ingredients",
+                placeholder="e.g., Arborio rice, black truffle dust, parmesan...",
+                lines=4
+            )
             btn = gr.Button("✨ UNLEASH MAGIC ✨", elem_classes="primary-btn")
 
         with gr.Column(elem_classes="group-container"):
             gr.Markdown("#### 🔮 THE REVELATION")
-            v_out = gr.Textbox(label="Vibe Detected", interactive=False, show_label=True)
+            v_out = gr.Textbox(
+                label="Vibe Detected",
+                interactive=False
+            )
             m_out = gr.Markdown()
-            # תיקון: הורדנו את show_download_button שגרם לשגיאה
-            im_out = gr.Image(label="AI Visualization", type="pil", interactive=False)
+            im_out = gr.Image(
+                label="AI Visualization",
+                type="pil",
+                interactive=False
+            )
 
     gr.Markdown("### ⚡ Instant Inspirations (Click to try):")
     gr.Examples(
         examples=[
-            ["Midnight Burger", "Black bun, wagyu beef, spicy aioli, caramelized onions"],
-            ["Neon Sushi Roll", "Tuna, salmon, avocado, glowing tobiko, eel sauce"],
-            ["Enchanted Forest Salad", "Kale, edible flowers, berries, nuts, fairy dust dressing"]
+            ["Midnight Burger",
+             "Black bun, wagyu beef, spicy aioli, caramelized onions"],
+            ["Neon Sushi Roll",
+             "Tuna, salmon, avocado, glowing tobiko, eel sauce"],
+            ["Enchanted Forest Salad",
+             "Kale, edible flowers, berries, nuts, fairy dust dressing"]
         ],
         inputs=[t_in, i_in],
         elem_id="examples-table"
@@ -158,4 +218,4 @@ with gr.Blocks(css=ultra_css, theme=theme) as demo:
     btn.click(smart_chef_app, [t_in, i_in], [v_out, m_out, im_out])
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()