Update app.py

app.py

@@ -212,15 +212,71 @@
     
 #     return "\n\n".join(results)
 
+# # ------------------------------
+# # Topic Validation (Check if topic belongs to subject)
+# # ------------------------------
+# def validate_topic_subject(topic, subject):
+#     """
+#     Validate if the topic belongs to the selected subject using LLM
+#     Returns True if valid, False otherwise
+#     """
+#     if not groq_client:
+#         return True  # Skip validation if API not available
+    
+#     validation_prompt = f"""You are a Class 12 PCB subject expert. Determine if the following topic belongs to {subject.title()}.
+
+# Topic: "{topic}"
+# Subject: {subject.title()}
+
+# Class 12 {subject.title()} covers:
+# {"- Reproduction, Genetics, Evolution, Plant Physiology, Human Systems, Ecology, Biotechnology" if subject == "biology" else ""}
+# {"- Solid State, Solutions, Thermodynamics, Electrochemistry, Organic Chemistry, Coordination Compounds" if subject == "chemistry" else ""}
+# {"- Rotational Dynamics, Fluids, Thermodynamics, Waves, Optics, Electromagnetism, Modern Physics, Semiconductors" if subject == "physics" else ""}
+
+# Answer ONLY with "YES" if the topic belongs to {subject.title()}, or "NO" if it belongs to a different subject.
+
+# Answer:"""
+    
+#     try:
+#         response = groq_client.chat.completions.create(
+#             messages=[
+#                 {
+#                     "role": "system",
+#                     "content": f"You are an expert at identifying which subject a topic belongs to. Answer only YES or NO."
+#                 },
+#                 {
+#                     "role": "user",
+#                     "content": validation_prompt
+#                 }
+#             ],
+#             model="llama-3.3-70b-versatile",
+#             temperature=0.1,
+#             max_tokens=10
+#         )
+        
+#         result = response.choices[0].message.content.strip().upper()
+        
+#         if "YES" in result:
+#             print(f"✓ Topic '{topic}' validated for {subject}")
+#             return True
+#         else:
+#             print(f"❌ Topic '{topic}' does NOT belong to {subject}")
+#             return False
+            
+#     except Exception as e:
+#         print(f"⚠️ Validation failed: {e}")
+#         return True  # Allow on error to avoid blocking
+
 # # ------------------------------
 # # Chapter Detection (Using Actual Chapter Names)
 # # ------------------------------
 # def detect_chapter_from_list(context, topic, subject):
 #     """
 #     Detect chapter using the actual chapter list by matching keywords
+#     Returns None if topic doesn't match the subject
 #     """
 #     if subject not in CHAPTER_NAMES:
-#         return "Unknown Chapter"
+#         return None
     
 #     chapters = CHAPTER_NAMES[subject]
 #     combined_text = (topic + " " + context[:1000]).lower()
@@ -258,32 +314,35 @@
 # def detect_chapter_with_llm(context, topic, subject, chapters):
 #     """
 #     Use LLM to pick the correct chapter from the provided list
+#     Also verifies if the topic belongs to the subject
 #     """
 #     if not groq_client:
-#         return "Unknown Chapter"
+#         return None
     
 #     chapter_list = "\n".join([f"{i+1}. {ch}" for i, ch in enumerate(chapters)])
     
-#     detection_prompt = f"""Based on the following textbook content and topic, identify which chapter from the list below this content belongs to.
+#     detection_prompt = f"""Based on the following textbook content and topic, identify which chapter from the Class 12 {subject.title()} textbook this content belongs to.
 
 # Topic: {topic}
 
 # Content snippet:
 # {context[:600]}
 
-# Available chapters:
+# Available {subject.title()} chapters:
 # {chapter_list}
 
-# Respond with ONLY the chapter number and name exactly as listed (e.g., "5. Origin and Evolution of Life"). Choose the most relevant chapter.
+# IMPORTANT: If the topic and content do NOT belong to {subject.title()}, respond with "NOT_MATCHING".
+
+# If it matches, respond with ONLY the chapter number and name exactly as listed (e.g., "5. Origin and Evolution of Life").
 
-# Chapter:"""
+# Response:"""
     
 #     try:
 #         response = groq_client.chat.completions.create(
 #             messages=[
 #                 {
 #                     "role": "system",
-#                     "content": "You are an expert at identifying which chapter textbook content belongs to. Respond with only the chapter number and name from the provided list."
+#                     "content": f"You are an expert at identifying which chapter textbook content belongs to. You can recognize when content doesn't match the subject. If the topic is from a different subject than {subject.title()}, respond with 'NOT_MATCHING'."
 #                 },
 #                 {
 #                     "role": "user",
@@ -297,6 +356,11 @@
         
 #         result = response.choices[0].message.content.strip()
         
+#         # Check if topic doesn't match the subject
+#         if "NOT_MATCHING" in result.upper() or "NOT MATCHING" in result.upper():
+#             print(f"⚠️ Topic '{topic}' doesn't belong to {subject}")
+#             return None
+        
 #         # Extract chapter name from response (remove number prefix if present)
 #         chapter = re.sub(r'^\d+\.\s*', '', result).strip()
         
@@ -307,11 +371,11 @@
 #                 return ch
         
 #         print(f"⚠️ LLM response not in list: {result}")
-#         return chapters[0]  # Default to first chapter
+#         return None
         
 #     except Exception as e:
 #         print(f"⚠️ Chapter detection failed: {e}")
-#         return "Unknown Chapter"
+#         return None
 
 # # ------------------------------
 # # MCQ Generation
@@ -325,7 +389,7 @@
 # 2. API key is valid (get one from https://console.groq.com/keys)
 # 3. Space has been restarted after adding the key
 # Current status: API key not found or invalid."""
-#         return error_msg, "Unknown"
+#         return error_msg, None
     
 #     # Check cache
 #     context_hash = hashlib.md5(context.encode()).hexdigest()[:8]
@@ -340,6 +404,12 @@
 #     # Detect the chapter from our actual chapter list
 #     chapter = detect_chapter_from_list(context, topic, subject)
     
+#     # If chapter is None, topic doesn't belong to this subject
+#     if chapter is None:
+#         error_msg = f"❌ The topic '{topic}' does not belong to {subject.title()}.\n\nPlease enter a topic related to {subject.title()} or select the correct subject."
+#         print(f"⚠️ Topic mismatch: '{topic}' not in {subject}")
+#         return error_msg, None
+    
 #     prompt = f"""You are a Class-12 {subject.title()} teacher creating MCQs.
 # Topic: "{topic}"
 # Chapter: "{chapter}"
@@ -723,8 +793,22 @@
 #         if subject not in SUBJECTS:
 #             return jsonify({"error": "Invalid subject"}), 400
         
-#         print(f"\n🔍 Searching {subject} for: {topic}")
+#         print(f"\n🔍 Validating topic for {subject}...")
+        
+#         # STEP 1: Validate if topic belongs to subject (BEFORE RAG search)
+#         if not validate_topic_subject(topic, subject):
+#             subject_names = {
+#                 "biology": "Biology",
+#                 "chemistry": "Chemistry", 
+#                 "physics": "Physics"
+#             }
+#             error_msg = f"The topic '{topic}' does not appear to be related to {subject_names[subject]}.\n\nPlease either:\n• Enter a {subject_names[subject]}-related topic, or\n• Select the correct subject for this topic"
+#             return jsonify({"error": error_msg}), 400
+        
+#         print(f"✓ Topic validated for {subject}")
+#         print(f"🔍 Searching {subject} for: {topic}")
         
+#         # STEP 2: RAG search
 #         context = rag_search(topic, subject, k=5)
         
 #         if not context or len(context.strip()) < 50:
@@ -732,8 +816,13 @@
         
 #         print(f"✓ Context found ({len(context)} chars)")
         
+#         # STEP 3: Generate MCQs
 #         mcqs, chapter = generate_mcqs(context, topic, subject)
         
+#         # Check if there was a subject mismatch
+#         if chapter is None:
+#             return jsonify({"error": mcqs}), 400
+        
 #         return jsonify({
 #             "mcqs": mcqs, 
 #             "subject": subject,
@@ -761,7 +850,6 @@
 #     port = int(os.environ.get("PORT", 7860))
 #     print(f"\n🚀 Starting server on port {port}...\n")
 #     app.run(host="0.0.0.0", port=port, debug=False)
-
 import pickle
 import faiss
 from flask import Flask, request, jsonify, render_template_string
@@ -1144,7 +1232,7 @@ Response:"""
 # ------------------------------
 # MCQ Generation
 # ------------------------------
-def generate_mcqs(context, topic, subject):
+def generate_mcqs(context, topic, subject, num_questions=5):
     # Check if Groq is available
     if not groq_client:
         error_msg = """ERROR: Groq API not initialized!
@@ -1157,13 +1245,13 @@ Current status: API key not found or invalid."""
     
     # Check cache
     context_hash = hashlib.md5(context.encode()).hexdigest()[:8]
-    cache_key = get_cache_key(topic, subject, context_hash)
+    cache_key = get_cache_key(topic, subject, context_hash) + f":{num_questions}"
     
     if cache_key in MCQ_CACHE:
         print("✓ Using cached MCQs")
         return MCQ_CACHE[cache_key]["mcqs"], MCQ_CACHE[cache_key]["chapter"]
     
-    print(f"🤖 Generating MCQs for {subject} - {topic}")
+    print(f"🤖 Generating {num_questions} MCQs for {subject} - {topic}")
     
     # Detect the chapter from our actual chapter list
     chapter = detect_chapter_from_list(context, topic, subject)
@@ -1180,7 +1268,7 @@ Chapter: "{chapter}"
 Reference material from textbook:
 {context[:1500]}
 
-Generate exactly 5 multiple-choice questions based on the reference material.
+Generate exactly {num_questions} multiple-choice questions based on the reference material.
 
 FORMAT (follow EXACTLY):
 Q1. [Question based on material]
@@ -1197,7 +1285,7 @@ C) [Option 3]
 D) [Option 4]
 Answer: [A/B/C/D] - [Brief explanation]
 
-Continue for Q3, Q4, Q5.
+Continue for Q3, Q4, Q5{"..." if num_questions > 5 else ""}.
 
 REQUIREMENTS:
 - All questions must be answerable from the reference material
@@ -1205,9 +1293,12 @@ REQUIREMENTS:
 - Correct answer must be clearly supported by material
 - Keep explanations brief (1-2 sentences)
 
-Generate 5 MCQs now:"""
+Generate {num_questions} MCQs now:"""
     
     try:
+        # Adjust max_tokens based on number of questions
+        max_tokens = min(3000, 300 * num_questions)
+        
         chat_completion = groq_client.chat.completions.create(
             messages=[
                 {
@@ -1221,7 +1312,7 @@ Generate 5 MCQs now:"""
             ],
             model="llama-3.3-70b-versatile",
             temperature=0.3,
-            max_tokens=1500,
+            max_tokens=max_tokens,
             top_p=0.9
         )
         
@@ -1449,7 +1540,33 @@ HTML_TEMPLATE = """
                 <input type="text" id="topic" placeholder="e.g., Mitochondria, Chemical Bonding, Newton's Laws">
             </div>
             
-            <button onclick="generateMCQs()">🚀 Generate 5 MCQs</button>
+            <div class="form-group">
+                <label for="numQuestions">🔢 Number of MCQs</label>
+                <select id="numQuestions">
+                    <option value="1">1 MCQ</option>
+                    <option value="2">2 MCQs</option>
+                    <option value="3">3 MCQs</option>
+                    <option value="4">4 MCQs</option>
+                    <option value="5" selected>5 MCQs</option>
+                    <option value="6">6 MCQs</option>
+                    <option value="7">7 MCQs</option>
+                    <option value="8">8 MCQs</option>
+                    <option value="9">9 MCQs</option>
+                    <option value="10">10 MCQs</option>
+                    <option value="11">11 MCQs</option>
+                    <option value="12">12 MCQs</option>
+                    <option value="13">13 MCQs</option>
+                    <option value="14">14 MCQs</option>
+                    <option value="15">15 MCQs</option>
+                    <option value="16">16 MCQs</option>
+                    <option value="17">17 MCQs</option>
+                    <option value="18">18 MCQs</option>
+                    <option value="19">19 MCQs</option>
+                    <option value="20">20 MCQs</option>
+                </select>
+            </div>
+            
+            <button onclick="generateMCQs()">🚀 Generate MCQs</button>
             
             <div class="loading" id="loading">
                 <div class="spinner"></div>
@@ -1479,6 +1596,7 @@ HTML_TEMPLATE = """
         async function generateMCQs() {
             const subject = document.getElementById('subject').value;
             const topic = document.getElementById('topic').value.trim();
+            const numQuestions = parseInt(document.getElementById('numQuestions').value);
             
             if (!topic) {
                 alert('⚠️ Please enter a topic!');
@@ -1500,7 +1618,7 @@ HTML_TEMPLATE = """
                 const response = await fetch('/generate', {
                     method: 'POST',
                     headers: {'Content-Type': 'application/json'},
-                    body: JSON.stringify({subject, topic})
+                    body: JSON.stringify({subject, topic, num_questions: numQuestions})
                 });
                 
                 const data = await response.json();
@@ -1523,7 +1641,7 @@ HTML_TEMPLATE = """
             } finally {
                 loading.classList.remove('show');
                 btn.disabled = false;
-                btn.textContent = '🚀 Generate 5 MCQs';
+                btn.textContent = '🚀 Generate MCQs';
             }
         }
         
@@ -1550,6 +1668,15 @@ def generate():
         data = request.json
         subject = data.get("subject", "").lower()
         topic = data.get("topic", "")
+        num_questions = data.get("num_questions", 5)
+        
+        # Validate num_questions
+        try:
+            num_questions = int(num_questions)
+            if num_questions < 1 or num_questions > 20:
+                num_questions = 5
+        except:
+            num_questions = 5
         
         if not topic:
             return jsonify({"error": "Topic is required"}), 400
@@ -1581,7 +1708,7 @@ def generate():
         print(f"✓ Context found ({len(context)} chars)")
         
         # STEP 3: Generate MCQs
-        mcqs, chapter = generate_mcqs(context, topic, subject)
+        mcqs, chapter = generate_mcqs(context, topic, subject, num_questions)
         
         # Check if there was a subject mismatch
         if chapter is None: