Update app.py

app.py

@@ -5,63 +5,59 @@ import faiss
 from sentence_transformers import SentenceTransformer
 import joblib
 
-# ===============================
-# 🔹 Load all assets
-# ===============================
+# Load assets
+print("🔄 Loading data and models...")
 df = pd.read_csv("clean_feedback.csv")
-embeddings = np.load("embeddings.npy")  # ✅ matches your file spelling
-index = faiss.read_index("feedback.index")  # ✅ matches the visible file
-clf = joblib.load("feedback_model.pkl")  # ✅ your classifier model
+print("✅ CSV loaded with columns:", df.columns.tolist())
+
+embeddings = np.load("embedings.npy")
+print("✅ Embeddings loaded with shape:", embeddings.shape)
+
+index = faiss.read_index("feedback.index")
+print("✅ FAISS index loaded")
+
+clf = joblib.load("feedback_model.pkl")
+print("✅ Sentiment model loaded")
 
-# ✅ Use CPU to avoid Hugging Face GPU issues
 model = SentenceTransformer("paraphrase-multilingual-MiniLM-L12-v2", device="cpu")
+print("✅ SentenceTransformer ready")
 
-# ===============================
-# 🔹 Define feedback classifier
-# ===============================
 def classify_feedback(text, top_k=5):
+    print(f"\n🧠 New query: {text}")
+
     if not text.strip():
-        return "**Please enter some feedback text.**"
+        return "⚠️ Please enter a feedback text."
 
-    # Create embedding for user input
+    # Embed query
     query_emb = model.encode([text])
+    print("Embedding shape:", query_emb.shape)
 
-    # Retrieve top-k similar feedbacks from FAISS index
+    # Search similar samples
     distances, indices = index.search(query_emb, top_k)
+    print("Retrieved indices:", indices)
 
-    # Get similar sentences
     retrieved = df.iloc[indices[0]]
+    if "Sentence" not in df.columns:
+        return "❌ Column 'Sentence' not found in CSV. Columns are: " + ", ".join(df.columns)
 
-    # Combine context (optional, for RAG explainability)
-    context = "\n".join(retrieved['Sentence'].astype(str).tolist())
-
-    # Predict sentiment from classifier
-    sentiment = clf.predict(query_emb)[0]
+    # Predict sentiment
+    try:
+        sentiment = clf.predict(query_emb)[0]
+    except Exception as e:
+        return f"❌ Model prediction error: {str(e)}"
 
-    # Prepare explanation list
-    examples = "\n".join(
-        [f"{i+1}. {s}" for i, s in enumerate(retrieved['Sentence'].astype(str).tolist())]
-    )
+    examples = "\n".join([f"{i+1}. {s}" for i, s in enumerate(retrieved['Sentence'].tolist())])
+    print("✅ Prediction done")
 
     return f"**Predicted Sentiment:** {sentiment}\n\n**Similar Feedbacks:**\n{examples}"
 
-# ===============================
-# 🔹 Gradio UI
-# ===============================
+# Gradio UI
 demo = gr.Interface(
     fn=classify_feedback,
-    inputs=[
-        gr.Textbox(
-            label="Enter Student Feedback",
-            placeholder="Type a Roman Urdu or English feedback sentence here..."
-        )
-    ],
+    inputs=[gr.Textbox(label="Enter Student Feedback")],
     outputs=[gr.Markdown(label="Prediction & Explanation")],
     title="🎓 Student Feedback RAG System",
-    description="Analyzes student feedback (Roman Urdu or English) using Retrieval-Augmented Generation with contextual examples."
+    description="Classifies Roman Urdu/English student feedback with context and reasoning."
 )
 
-# ===============================
-# 🔹 Launch app
-# ===============================
 demo.launch()