Update app.py

app.py

@@ -4,240 +4,207 @@ from datasets import load_dataset
 import chromadb
 from chromadb.config import Settings
 
+print("="*70)
 print("Loading Restaurant Review Advisor...")
+print("="*70)
 
 # ============================================================================
-# COMPONENT 1: LOAD SENTIMENT MODEL FROM HUGGING FACE
+# COMPONENT 1: LOAD SENTIMENT MODEL
 # ============================================================================
-print("[1/3] Loading fine-tuned sentiment model...")
+print("\n[1/3] Loading fine-tuned sentiment model...")
 
-# Load YOUR trained model from Hugging Face
-# Replace 'Isap31' with your actual username if different
-sentiment_analyzer = pipeline(
-    "sentiment-analysis",
-    model="Isap31/restaurant-sentiment-distilbert"
-)
-
-print("✅ Sentiment model loaded (94.93% accuracy)")
+try:
+    sentiment_analyzer = pipeline(
+        "sentiment-analysis",
+        model="Isap31/restaurant-sentiment-distilbert"
+    )
+    print("✅ Sentiment model loaded (94.93% accuracy)")
+except Exception as e:
+    print(f"Error loading sentiment model: {e}")
+    sentiment_analyzer = None
 
 # ============================================================================
 # COMPONENT 2: SETUP RAG SYSTEM
 # ============================================================================
-print("[2/3] Setting up RAG knowledge base...")
-
-# Load Yelp reviews dataset
-dataset = load_dataset("fancyzhx/yelp_polarity", split="train")
-sampled_reviews = dataset.shuffle(seed=42).select(range(500))  # 2000 for faster loading
-
-# Create ChromaDB collection
-chroma_client = chromadb.Client(Settings(
-    anonymized_telemetry=False,
-    allow_reset=True
-))
-
-collection = chroma_client.create_collection(name="yelp_reviews")
-
-# Add reviews to vector database
-documents = []
-metadatas = []
-ids = []
-
-for idx, review in enumerate(sampled_reviews):
-    if len(review['text']) >= 50:
-        documents.append(review['text'])
-        metadatas.append({
-            'sentiment': 'positive' if review['label'] == 1 else 'negative'
-        })
-        ids.append(f"review_{idx}")
-
-# Add to collection
-collection.add(
-    documents=documents,
-    metadatas=metadatas,
-    ids=ids
-)
-
-print(f"✅ RAG knowledge base ready with {len(documents)} reviews")
-
-# Load generation model for RAG
-generator = pipeline(
-    "text2text-generation",
-    model="google/flan-t5-small",
-    max_length=150
-)
+print("\n[2/3] Setting up RAG knowledge base...")
 
-print("✅ RAG generation model loaded")
+try:
+    # Load dataset
+    dataset = load_dataset("fancyzhx/yelp_polarity", split="train")
+    sampled_reviews = dataset.shuffle(seed=42).select(range(500))  # Reduced for stability
+    
+    # Create ChromaDB
+    chroma_client = chromadb.Client(Settings(
+        anonymized_telemetry=False,
+        allow_reset=True
+    ))
+    
+    collection = chroma_client.create_collection(name="yelp_reviews")
+    
+    # Add reviews
+    documents = []
+    metadatas = []
+    ids = []
+    
+    for idx, review in enumerate(sampled_reviews):
+        if len(review['text']) >= 50:
+            documents.append(review['text'])
+            metadatas.append({'sentiment': 'positive' if review['label'] == 1 else 'negative'})
+            ids.append(f"review_{idx}")
+    
+    collection.add(documents=documents, metadatas=metadatas, ids=ids)
+    
+    print(f"✅ RAG knowledge base ready with {len(documents)} reviews")
+    
+    # Load generation model
+    generator = pipeline("text2text-generation", model="google/flan-t5-small", max_length=150)
+    print("✅ RAG generation model loaded")
+    
+    rag_ready = True
+except Exception as e:
+    print(f"Error setting up RAG: {e}")
+    rag_ready = False
+    collection = None
+    generator = None
 
 # ============================================================================
-# DEFINE APPLICATION FUNCTIONS
+# DEFINE FUNCTIONS
 # ============================================================================
-print("[3/3] Setting up Gradio interface...")
+print("\n[3/3] Setting up application...")
 
 def analyze_sentiment(text):
     """Analyze sentiment of restaurant review"""
     if not text.strip():
-        return "⚪ No input", "N/A", ""
+        return "⚪ No input provided"
     
-    result = sentiment_analyzer(text)[0]
-    label = result['label']
-    confidence = result['score']
+    if sentiment_analyzer is None:
+        return "❌ Sentiment model not loaded"
     
-    if label.upper() in ['POSITIVE', 'LABEL_1', '1']:
-        return "🟢 POSITIVE", f"{confidence:.1%}", "Great review! Customer is satisfied."
-    else:
-        return "🔴 NEGATIVE", f"{confidence:.1%}", "Negative feedback detected."
+    try:
+        result = sentiment_analyzer(text)[0]
+        label = result['label']
+        confidence = result['score']
+        
+        if label.upper() in ['POSITIVE', 'LABEL_1', '1']:
+            sentiment = f"🟢 POSITIVE ({confidence:.1%} confidence)"
+            interpretation = "Great review! Customer is satisfied."
+        else:
+            sentiment = f"🔴 NEGATIVE ({confidence:.1%} confidence)"
+            interpretation = "Negative feedback detected."
+        
+        return f"{sentiment}\n\n{interpretation}"
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
 
 def rag_query(question):
     """RAG system - retrieve reviews and generate answer"""
     if not question.strip():
         return "Please ask a question about restaurants."
     
-    # Retrieval
-    results = collection.query(query_texts=[question], n_results=3)
-    
-    if not results['documents'][0]:
-        return "I couldn't find relevant reviews for that question."
-    
-    # Build context
-    review_texts = []
-    sentiments = []
-    
-    for doc, metadata in zip(results['documents'][0], results['metadatas'][0]):
-        sentiment = metadata.get('sentiment', 'unknown')
-        sentiments.append(sentiment)
-        review_texts.append(f"[{sentiment.upper()}] {doc}")
-    
-    context = "\n\n".join(review_texts)
+    if not rag_ready:
+        return "❌ RAG system not loaded"
     
-    # Augmentation & Generation
-    prompt = f"""Based on these customer reviews, answer the question concisely.
+    try:
+        # Retrieval
+        results = collection.query(query_texts=[question], n_results=3)
+        
+        if not results['documents'][0]:
+            return "I couldn't find relevant reviews for that question."
+        
+        # Build context
+        review_texts = []
+        sentiments = []
+        
+        for doc, metadata in zip(results['documents'][0], results['metadatas'][0]):
+            sentiment = metadata.get('sentiment', 'unknown')
+            sentiments.append(sentiment)
+            review_texts.append(f"[{sentiment.upper()}] {doc}")
+        
+        context = "\n\n".join(review_texts)
+        
+        # Generate
+        prompt = f"""Based on these customer reviews, answer the question concisely.
+
 Reviews:
 {context}
+
 Question: {question}
-Answer:"""
-    
-    answer = generator(prompt, max_length=150)[0]['generated_text']
-    
-    # Format response
-    response = f"**Generated Answer:**\n{answer}\n\n"
-    response += f"**Based on:** {len(results['documents'][0])} customer reviews "
-    response += f"({sentiments.count('positive')} positive, {sentiments.count('negative')} negative)"
-    
-    return response
 
-def complete_advisor(user_input, mode):
-    """Main function - routes to appropriate component"""
-    if not user_input.strip():
-        return "Please enter text or a question."
-    
-    if mode == "Sentiment Analysis":
-        sentiment, confidence, interpretation = analyze_sentiment(user_input)
-        return f"**Sentiment:** {sentiment}\n**Confidence:** {confidence}\n\n{interpretation}"
-    
-    elif mode == "Knowledge Query (RAG)":
-        return rag_query(user_input)
-    
-    else:  # Complete Analysis
-        sentiment, confidence, interpretation = analyze_sentiment(user_input)
-        rag_answer = rag_query(user_input)
+Answer:"""
+        
+        answer = generator(prompt, max_length=150)[0]['generated_text']
+        
+        # Format response
+        response = f"**Generated Answer:**\n{answer}\n\n"
+        response += f"**Based on:** {len(results['documents'][0])} customer reviews "
+        response += f"({sentiments.count('positive')} positive, {sentiments.count('negative')} negative)"
         
-        result = f"## 📊 Sentiment Analysis\n"
-        result += f"**Result:** {sentiment}\n"
-        result += f"**Confidence:** {confidence}\n"
-        result += f"**Interpretation:** {interpretation}\n\n"
-        result += f"## 🤖 RAG Knowledge System\n\n{rag_answer}"
+        return response
         
-        return result
+    except Exception as e:
+        return f"❌ Error: {str(e)}"
 
 # ============================================================================
 # CREATE GRADIO INTERFACE
 # ============================================================================
+print("\nCreating Gradio interface...")
+
+# Create the interface
+demo = gr.Interface(
+    fn=analyze_sentiment,
+    inputs=gr.Textbox(
+        lines=3,
+        placeholder="Enter restaurant review...",
+        label="Review Text"
+    ),
+    outputs=gr.Textbox(label="Sentiment Analysis Result"),
+    title="🍽️ Restaurant Review Advisor",
+    description="""
+    **Fine-Tuned Sentiment Analysis + RAG Knowledge System**
+    
+    Component 1: DistilBERT sentiment model (94.93% accuracy)
+    Component 2: RAG system with 500+ Yelp reviews
+    Component 3: Gradio interface
+    
+    Enter a restaurant review to analyze its sentiment!
+    """,
+    examples=[
+        ["This restaurant exceeded all expectations! The service was impeccable and food was divine."],
+        ["Worst dining experience ever. Cold food, rude staff, and overpriced."],
+        ["The ambiance was nice, but the food was just okay."],
+    ],
+    theme=gr.themes.Soft()
+)
 
-with gr.Blocks(theme=gr.themes.Soft(), title="Restaurant Review Advisor") as demo:
-    gr.HTML('''
-        <h1 style="text-align: center; background: linear-gradient(90deg, #f093fb 0%, #f5576c 100%);
-        -webkit-background-clip: text; -webkit-text-fill-color: transparent; font-size: 3em;">
-        ☕ Restaurant Review Advisor
-        </h1>
-    ''')
-    gr.HTML('<p style="text-align: center; font-size: 1.3em; color: #666;">Fine-Tuned Sentiment Analysis + RAG Knowledge System</p>')
-    
-    with gr.Row():
-        with gr.Column(scale=2):
-            gr.Markdown("""
-            ### 🎯 System Overview
-            
-            **Component 1:** Fine-tuned DistilBERT sentiment model
-            - Trained on 50,000 Yelp reviews
-            - **Accuracy: 94.93%**
-            - Binary classification (Positive/Negative)
-            
-            **Component 2:** RAG System with Yelp Reviews
-            - Retrieval: ChromaDB vector search
-            - Documents: 2,000+ real customer reviews
-            - Generation: FLAN-T5 language model
-            
-            **Component 3:** Integrated Gradio Interface
-            - Three analysis modes
-            - Real-time processing
-            - Public deployment
-            """)
-    
-    gr.Markdown("---")
-    
-    with gr.Row():
-        with gr.Column():
-            mode = gr.Radio(
-                choices=["Sentiment Analysis", "Knowledge Query (RAG)", "Complete Analysis"],
-                value="Sentiment Analysis",
-                label="🎛️ Select Mode"
-            )
-            
-            user_input = gr.Textbox(
-                lines=4,
-                placeholder="Enter restaurant review or question...",
-                label="📝 Input"
-            )
-            
-            submit_btn = gr.Button("🚀 Analyze", variant="primary", size="lg")
-        
-        with gr.Column():
-            output = gr.Markdown(label="📊 Results")
-    
-    gr.Markdown("---")
-    gr.Markdown("### 💡 Try These Examples:")
-    
-    gr.Examples(
-        examples=[
-            ["This restaurant exceeded all expectations! The service was impeccable and food was divine.", "Sentiment Analysis"],
-            ["Worst dining experience ever. Cold food, rude staff, and overpriced.", "Sentiment Analysis"],
-            ["What do customers say about food quality at restaurants?", "Knowledge Query (RAG)"],
-            ["Are portions typically good at restaurants?", "Knowledge Query (RAG)"],
-            ["The ambiance was perfect! What do people say about restaurant atmosphere?", "Complete Analysis"],
-        ],
-        inputs=[user_input, mode]
-    )
-    
-    submit_btn.click(
-        fn=complete_advisor,
-        inputs=[user_input, mode],
-        outputs=output
-    )
-    
-    gr.Markdown("""
-    ---
-    ### 📚 Technical Details
-    
-    **Model:** [Isap31/restaurant-sentiment-distilbert](https://huggingface.co/Isap31/restaurant-sentiment-distilbert)  
-    **Dataset:** [Yelp Review Polarity](https://huggingface.co/datasets/fancyzhx/yelp_polarity)  
-    **Framework:** Hugging Face Transformers, ChromaDB, Gradio  
-    **GitHub:** [Restaurant Review Advisor](https://github.com/Isap31/Restaurant-review-advisor-Final-Project-452)
-    
-    **INFO 452 Final Project** | Fall 2025 | VCU
-    """)
+# Create RAG interface
+demo_rag = gr.Interface(
+    fn=rag_query,
+    inputs=gr.Textbox(
+        lines=2,
+        placeholder="Ask a question about restaurants...",
+        label="Question"
+    ),
+    outputs=gr.Textbox(label="RAG Answer"),
+    title="🔍 Restaurant Knowledge Query (RAG)",
+    description="Ask questions about restaurants and get answers based on real customer reviews!",
+    examples=[
+        ["What do customers say about food quality at restaurants?"],
+        ["Are portions typically good at restaurants?"],
+        ["What about customer service?"],
+    ],
+    theme=gr.themes.Soft()
+)
+
+# Combine with tabs
+app = gr.TabbedInterface(
+    [demo, demo_rag],
+    ["Sentiment Analysis", "RAG Knowledge Query"],
+    title="☕ Restaurant Review Advisor - INFO 452 Final Project"
+)
 
 print("✅ Gradio interface ready!")
+print("="*70)
 
 # Launch
 if __name__ == "__main__":
-    demo.launch()
+    app.launch()