fix: resolve ambiguous column error and improve AI suggestions UX

- Fix ambiguous 'session_id' in classifier_routes.py SQL query.
- Update nvidia_prompts.py to return multiple chapter suggestions and remove redundant text field.
- Enhance question_entry_v2.html with suggestion pre-fetching, caching, and multi-choice chip UI.

classifier_routes.py

@@ -43,7 +43,7 @@ def get_topic_suggestions():
         try:
             conn = get_db_connection()
             # Check DB first
-            row = conn.execute('SELECT question_text, processed_filename, session_id FROM questions q JOIN images i ON q.image_id = i.id WHERE i.id = ?', (image_id,)).fetchone()
+            row = conn.execute('SELECT question_text, processed_filename, i.session_id FROM questions q JOIN images i ON q.image_id = i.id WHERE i.id = ?', (image_id,)).fetchone()
             
             if row:
                 if row['question_text']:
@@ -101,7 +101,7 @@ def get_topic_suggestions():
                 "role": "user"
             }
         ],
-        "temperature": 0.1, # Low temp for deterministic classification
+        "temperature": 0.2, # Slightly higher for variety in top-k if supported, but here we just want accurate multiple suggestions
         "top_p": 1,
         "max_tokens": 1024,
         "stream": False
@@ -122,12 +122,19 @@ def get_topic_suggestions():
              
         data = json.loads(content)
         
-        # Extract the chapter title
-        chapter_title = "Unclassified"
+        # Extract suggestions
+        suggestions = []
         if data.get('data') and len(data['data']) > 0:
-            chapter_title = data['data'][0].get('chapter_title', 'Unclassified')
+            primary_chapter = data['data'][0].get('chapter_title', 'Unclassified')
+            suggestions.append(primary_chapter)
             
-        return jsonify({'success': True, 'chapter_title': chapter_title, 'full_response': data})
+            # Check for alternative suggestions if the model provides them (we will update prompts to support this)
+            if 'other_possible_chapters' in data['data'][0]:
+                others = data['data'][0]['other_possible_chapters']
+                if isinstance(others, list):
+                    suggestions.extend(others)
+            
+        return jsonify({'success': True, 'suggestions': suggestions, 'full_response': data})
 
     except Exception as e:
         current_app.logger.error(f"NVIDIA API Error: {e}")

nvidia_prompts.py

@@ -1,4 +1,3 @@
-
 BIOLOGY_PROMPT_TEMPLATE = """**System Role:** You are a specialized Biology question classifier for NEET exams. Your task is to map questions to their corresponding chapter from the NCERT Biology syllabus.
 
 **Syllabus Chapters (Use these exact titles):**
@@ -43,7 +42,8 @@ BIOLOGY_PROMPT_TEMPLATE = """**System Role:** You are a specialized Biology ques
 
 1. **Scope**: Analyze the input question. If the question is NOT related to Biology, mark the chapter as 'Unclassified'.
 2. **Mapping**: Identify the single most relevant chapter from the list above.
-3. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include both.
+3. **Ambiguity**: If other chapters are also plausible, list up to 2 alternatives in 'other_possible_chapters'.
+4. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include the primary one in 'chapter_title' and the secondary in 'other_possible_chapters'.
 
 **Output JSON Schema:**
 
@@ -55,7 +55,7 @@ BIOLOGY_PROMPT_TEMPLATE = """**System Role:** You are a specialized Biology ques
       "subject": "Biology",
       "chapter_index": <chapter number or 0>,
       "chapter_title": "<exact chapter title from list or 'Unclassified'>",
-      "original_question_text": "<complete original question>",
+      "other_possible_chapters": ["<alternative chapter 1>", "<alternative chapter 2>"],
       "confidence": <0.0 to 1.0>
     }
   ],
@@ -113,7 +113,8 @@ syllabus.
 
 1. **Scope**: Analyze the input question. If the question is NOT related to Chemistry, mark the chapter as 'Unclassified'.
 2. **Mapping**: Identify the single most relevant chapter from the list above.
-3. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include both.
+3. **Ambiguity**: If other chapters are also plausible, list up to 2 alternatives in 'other_possible_chapters'.
+4. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include the primary one in 'chapter_title' and the secondary in 'other_possible_chapters'.
 
 **Output JSON Schema:**
 
@@ -125,7 +126,7 @@ syllabus.
       "subject": "Chemistry",
       "chapter_index": <chapter number or 0>,
       "chapter_title": "<exact chapter title from list or 'Unclassified'>",
-      "original_question_text": "<complete original question>",
+      "other_possible_chapters": ["<alternative chapter 1>", "<alternative chapter 2>"],
       "confidence": <0.0 to 1.0>
     }
   ],
@@ -181,7 +182,8 @@ PHYSICS_PROMPT_TEMPLATE = """**System Role:** You are a specialized Physics ques
 
 1. **Scope**: Analyze the input question. If the question is NOT related to Physics, mark the chapter as 'Unclassified'.
 2. **Mapping**: Identify the single most relevant chapter from the list above.
-3. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include both.
+3. **Ambiguity**: If other chapters are also plausible, list up to 2 alternatives in 'other_possible_chapters'.
+4. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include the primary one in 'chapter_title' and the secondary in 'other_possible_chapters'.
 
 **Output JSON Schema:**
 
@@ -193,7 +195,7 @@ PHYSICS_PROMPT_TEMPLATE = """**System Role:** You are a specialized Physics ques
       "subject": "Physics",
       "chapter_index": <chapter number or 0>,
       "chapter_title": "<exact chapter title from list or 'Unclassified'>",
-      "original_question_text": "<complete original question>",
+      "other_possible_chapters": ["<alternative chapter 1>", "<alternative chapter 2>"],
       "confidence": <0.0 to 1.0>
     }
   ],
@@ -250,7 +252,8 @@ syllabus.
 
 1. **Scope**: Analyze the input question. If the question is NOT related to Mathematics, mark the chapter as 'Unclassified'.
 2. **Mapping**: Identify the single most relevant chapter from the list above.
-3. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include both.
+3. **Ambiguity**: If other chapters are also plausible, list up to 2 alternatives in 'other_possible_chapters'.
+4. **Multi-Chapter**: If a question explicitly spans 2 distinct chapters, include the primary one in 'chapter_title' and the secondary in 'other_possible_chapters'.
 
 **Output JSON Schema:**
 
@@ -262,7 +265,7 @@ syllabus.
       "subject": "Mathematics",
       "chapter_index": <chapter number or 0>,
       "chapter_title": "<exact chapter title from list or 'Unclassified'>",
-      "original_question_text": "<complete original question>",
+      "other_possible_chapters": ["<alternative chapter 1>", "<alternative chapter 2>"],
       "confidence": <0.0 to 1.0>
     }
   ],
@@ -273,4 +276,4 @@ syllabus.
 
 **Input Questions:**
 {input_questions}
-"""
+"""

templates/question_entry_v2.html

@@ -330,7 +330,11 @@
                         </button>
                     </div>
                 </div>
-                <div id="ai_suggestion_result" class="alert alert-info d-none mt-2"></div>
+                <div id="ai_suggestion_container" class="d-none mt-2">
+                    <label class="form-label text-muted small">AI Suggestions:</label>
+                    <div id="ai_suggestion_chips" class="d-flex flex-wrap gap-2"></div>
+                </div>
+                <div id="ai_error_msg" class="alert alert-danger d-none mt-2"></div>
             </div>
             <div class="modal-footer d-flex justify-content-between">
                 <button type="button" class="btn btn-secondary" onclick="prevTopicQuestion()">Previous</button>
@@ -978,6 +982,7 @@
     let manualQuestionsList = [];
     let currentManualIndex = 0;
     let manualSubject = "";
+    let suggestionCache = new Map(); // Stores promises or results: imageId -> {suggestions: [], status: 'pending'|'done'}
 
     function parseRange(rangeStr, maxVal) {
         const indices = new Set();
@@ -995,7 +1000,6 @@
                 if (!isNaN(num)) indices.add(num);
             }
         }
-        // Filter valid 1-based indices and convert to 0-based
         return Array.from(indices).filter(i => i >= 1 && i <= maxVal).map(i => i - 1).sort((a, b) => a - b);
     }
 
@@ -1015,6 +1019,8 @@
         }
 
         currentManualIndex = 0;
+        suggestionCache.clear(); // Clear cache on new run
+
         const modal1 = bootstrap.Modal.getInstance(document.getElementById('manualClassificationModal'));
         modal1.hide();
         
@@ -1022,18 +1028,48 @@
         modal2.show();
         
         loadTopicQuestion();
+        
+        // Trigger prefetch for the first few questions
+        prefetchSuggestions(0, 2);
+    }
+
+    async function prefetchSuggestions(startIndex, count) {
+        for (let i = startIndex; i < startIndex + count && i < manualQuestionsList.length; i++) {
+            const globalIndex = manualQuestionsList[i];
+            const fieldset = document.querySelector(`fieldset[data-question-index="${globalIndex}"]`);
+            const imageId = fieldset.querySelector('input[name^="image_id_"]').value;
+            
+            if (!suggestionCache.has(imageId)) {
+                // Initiate fetch and store the promise
+                const promise = fetch('/get_topic_suggestions', {
+                    method: 'POST',
+                    headers: { 'Content-Type': 'application/json' },
+                    body: JSON.stringify({ image_id: imageId, subject: manualSubject })
+                })
+                .then(res => res.json())
+                .then(data => {
+                    if (data.success) {
+                        return { status: 'done', suggestions: data.suggestions || [data.chapter_title] };
+                    } else {
+                        throw new Error(data.error);
+                    }
+                })
+                .catch(err => ({ status: 'error', error: err.message }));
+                
+                suggestionCache.set(imageId, promise);
+            }
+        }
     }
 
     function loadTopicQuestion() {
         const globalIndex = manualQuestionsList[currentManualIndex];
-        // Find the fieldset for this index
         const fieldset = document.querySelector(`fieldset[data-question-index="${globalIndex}"]`);
         const imageId = fieldset.querySelector('input[name^="image_id_"]').value;
         const qNum = fieldset.querySelector('input[name^="question_number_"]').value;
         const img = fieldset.querySelector('img');
         const imgUrl = img ? img.src : '';
         
-        // Populate Modal
+        // UI Updates
         document.getElementById('topic_q_num').innerText = qNum;
         const modalImg = document.getElementById('topic_q_img');
         if (imgUrl) {
@@ -1045,31 +1081,49 @@
 
         document.getElementById('topic_subject_display').innerText = manualSubject;
         
-        // Get existing chapter if any
         const chapterSpan = document.getElementById(`chapter_${imageId}`);
         const currentChapter = chapterSpan ? chapterSpan.innerText : '';
         document.getElementById('topic_input').value = (currentChapter && currentChapter !== 'N/A' && currentChapter !== 'Unclassified') ? currentChapter : '';
         
         document.getElementById('topic_progress').innerText = `${currentManualIndex + 1} / ${manualQuestionsList.length}`;
-        const resultDiv = document.getElementById('ai_suggestion_result');
-        resultDiv.classList.add('d-none');
-        resultDiv.innerText = '';
         
-        // Reset button state
+        // Clear previous suggestions
+        document.getElementById('ai_suggestion_container').classList.add('d-none');
+        document.getElementById('ai_suggestion_chips').innerHTML = '';
+        document.getElementById('ai_error_msg').classList.add('d-none');
+        
+        // Reset button
         const btn = document.getElementById('btn_get_suggestion');
         btn.disabled = false;
         btn.innerHTML = `<span class="spinner-border spinner-border-sm d-none" role="status"></span> ✨ Get AI Suggestion`;
+
+        // Check cache for instant load (optional: auto-show if cached?)
+        // Let's stick to user clicking for now, or we can auto-load.
+        // The user asked "auto get multiple ai suggestions". So we should try to show them if available.
+        if (suggestionCache.has(imageId)) {
+             // If we have it (or it's loading), we can try to show it automatically or just wait for click.
+             // Given "auto get", let's simulate a click if the user hasn't entered a topic yet.
+             if (!document.getElementById('topic_input').value) {
+                 getAiSuggestion();
+             }
+        } else {
+            // Not in cache, start fetching now (prefetch logic handles this, but ensuring current is fetched)
+            prefetchSuggestions(currentManualIndex, 1);
+            if (!document.getElementById('topic_input').value) {
+                 getAiSuggestion();
+            }
+        }
     }
 
     async function nextTopicQuestion() {
-        // Save current
         await saveCurrentTopic();
         
         if (currentManualIndex < manualQuestionsList.length - 1) {
             currentManualIndex++;
             loadTopicQuestion();
+            // Prefetch next few
+            prefetchSuggestions(currentManualIndex + 1, 2);
         } else {
-            // Finish
             const modal2 = bootstrap.Modal.getInstance(document.getElementById('topicSelectionModal'));
             modal2.hide();
             showStatus("Manual classification completed for selected range.", "success");
@@ -1089,14 +1143,12 @@
         const imageId = fieldset.querySelector('input[name^="image_id_"]').value;
         const topic = document.getElementById('topic_input').value;
 
-        // Update UI immediately
         const subjectSpan = document.getElementById(`subject_${imageId}`);
         if(subjectSpan) subjectSpan.innerText = manualSubject;
         
         const chapterSpan = document.getElementById(`chapter_${imageId}`);
         if(chapterSpan) chapterSpan.innerText = topic || 'Unclassified';
 
-        // Send to backend
         try {
             await fetch('/classified/update_single', {
                 method: 'POST',
@@ -1117,34 +1169,62 @@
         const fieldset = document.querySelector(`fieldset[data-question-index="${globalIndex}"]`);
         const imageId = fieldset.querySelector('input[name^="image_id_"]').value;
         const btn = document.getElementById('btn_get_suggestion');
-        const resultDiv = document.getElementById('ai_suggestion_result');
+        const container = document.getElementById('ai_suggestion_container');
+        const chipsContainer = document.getElementById('ai_suggestion_chips');
+        const errorDiv = document.getElementById('ai_error_msg');
         
         btn.disabled = true;
         btn.querySelector('.spinner-border').classList.remove('d-none');
+        errorDiv.classList.add('d-none');
 
         try {
-            const response = await fetch('/get_topic_suggestions', {
-                method: 'POST',
-                headers: { 'Content-Type': 'application/json' },
-                body: JSON.stringify({
-                    image_id: imageId,
-                    subject: manualSubject
+            let result;
+            if (suggestionCache.has(imageId)) {
+                result = await suggestionCache.get(imageId);
+            } else {
+                // Should be covered by prefetch, but fallback just in case
+                const promise = fetch('/get_topic_suggestions', {
+                    method: 'POST',
+                    headers: { 'Content-Type': 'application/json' },
+                    body: JSON.stringify({ image_id: imageId, subject: manualSubject })
                 })
-            });
-            const data = await response.json();
-            
-            if (data.success) {
-                document.getElementById('topic_input').value = data.chapter_title;
-                resultDiv.innerText = `AI Suggestion: ${data.chapter_title}`;
-                resultDiv.classList.remove('d-none', 'alert-danger');
-                resultDiv.classList.add('alert-info');
+                .then(res => res.json())
+                .then(data => {
+                    if (data.success) return { status: 'done', suggestions: data.suggestions || [data.chapter_title] };
+                    else throw new Error(data.error);
+                })
+                .catch(err => ({ status: 'error', error: err.message }));
+                
+                suggestionCache.set(imageId, promise);
+                result = await promise;
+            }
+
+            if (result.status === 'done') {
+                container.classList.remove('d-none');
+                chipsContainer.innerHTML = '';
+                
+                result.suggestions.forEach(suggestion => {
+                    const chip = document.createElement('button');
+                    chip.className = 'btn btn-outline-info btn-sm rounded-pill';
+                    chip.innerText = suggestion;
+                    chip.onclick = () => {
+                        document.getElementById('topic_input').value = suggestion;
+                    };
+                    chipsContainer.appendChild(chip);
+                });
+                
+                // Auto-fill input if empty and we have a primary suggestion
+                const currentVal = document.getElementById('topic_input').value;
+                if (!currentVal && result.suggestions.length > 0) {
+                     document.getElementById('topic_input').value = result.suggestions[0];
+                }
+
             } else {
-                throw new Error(data.error);
+                throw new Error(result.error);
             }
         } catch (e) {
-            resultDiv.innerText = `Error: ${e.message}`;
-            resultDiv.classList.remove('d-none', 'alert-info');
-            resultDiv.classList.add('alert-danger');
+            errorDiv.innerText = `Error: ${e.message}`;
+            errorDiv.classList.remove('d-none');
         } finally {
             btn.disabled = false;
             btn.querySelector('.spinner-border').classList.add('d-none');