Update app.py

app.py

@@ -2,9 +2,15 @@ import streamlit as st
 import pandas as pd
 import re
 import string
+import google.generativeai as genai
 from sklearn.model_selection import train_test_split
 from sklearn.feature_extraction.text import TfidfVectorizer
 from sklearn.naive_bayes import MultinomialNB
+from sklearn.metrics.pairwise import cosine_similarity
+
+# --- Set Gemini API Key ---
+genai.configure(api_key="AIzaSyCVRGVxIe1vESoAgykgHWOej-jZxiU-RKE")  # <-- Replace this with your actual Gemini API key
+gemini_model = genai.GenerativeModel("gemini-pro")
 
 # Title & Intro
 st.set_page_config(page_title="SMS Spam Detection", layout="centered")
@@ -27,7 +33,6 @@ df['label'] = df['label'].map({'ham': 0, 'spam': 1})
 
 # --- Train Model ---
 X_train, X_test, y_train, y_test = train_test_split(df['message'], df['label'], test_size=0.2, random_state=42)
-
 vectorizer = TfidfVectorizer()
 X_train_tfidf = vectorizer.fit_transform(X_train)
 
@@ -51,6 +56,18 @@ def predict_spam(text):
     prediction = model.predict(vector)
     return "Spam" if prediction[0] == 1 else "Not Spam (Ham)"
 
+# --- Gemini Fallback ---
+def ask_gemini(text):
+    prompt = f"""You are an expert SMS spam detector.
+Classify the following message as 'Spam' or 'Not Spam (Ham)'.
+Message: "{text}"
+Reply with only: Spam or Not Spam (Ham)."""
+    try:
+        response = gemini_model.generate_content(prompt)
+        return response.text.strip()
+    except Exception as e:
+        return f"Error using Gemini: {str(e)}"
+
 # --- Input ---
 user_input = st.text_area("✉️ Enter your SMS message here:")
 
@@ -58,11 +75,25 @@ if st.button("Check Message"):
     if user_input.strip() == "":
         st.warning("⚠️ Please enter a message.")
     else:
-        result = predict_spam(user_input)
-        if result == "Spam":
-            st.error("🚫 This message is classified as **SPAM**.")
+        cleaned = clean_text(user_input)
+        input_vector = vectorizer.transform([cleaned])
+        similarities = cosine_similarity(input_vector, X_train_tfidf)
+        max_similarity = similarities.max()
+
+        # Check similarity threshold (e.g., < 0.3 = unknown message)
+        if max_similarity < 0.3:
+            st.info("🧠 Message not found in training data. Using Gemini for prediction...")
+            gemini_result = ask_gemini(user_input)
+            if "spam" in gemini_result.lower():
+                st.error("🚫 Gemini says: This message is **SPAM**.")
+            else:
+                st.success("✅ Gemini says: This message is **NOT SPAM (HAM)**.")
         else:
-            st.success("✅ This message is classified as **NOT SPAM (HAM)**.")
+            result = predict_spam(user_input)
+            if result == "Spam":
+                st.error("🚫 This message is classified as **SPAM**.")
+            else:
+                st.success("✅ This message is classified as **NOT SPAM (HAM)**.")
 
 # --- Dataset preview ---
 with st.expander("📄 View sample dataset"):