Roll back interruption change

app.py

@@ -10,7 +10,6 @@ from utils.models import generate_summaries, model_names
 from utils.ui_helpers import toggle_context_display, update_feedback, get_context_html, toggle_reference_answer
 from utils.leaderboard import load_leaderboard_data, submit_vote_with_elo, generate_leaderboard_html
 from utils.vote_logger import save_vote_details
-from utils.shared import generation_interrupt
 
 feedback_options = {
     "left": [
@@ -99,13 +98,8 @@ def weighted_sample_without_replacement(population, weights, k=2):
         
     return selected
 
-def load_context(set_interrupt=False):
-    if set_interrupt:
-        generation_interrupt.set()
-        time.sleep(0.5)  # Give more time for interrupt to take effect
-        
-    # DON'T clear the interrupt here - let the new inference clear it
-    # generation_interrupt.clear()  # REMOVED THIS LINE
+def load_context():
+    # Simplified - no interrupt logic
     example = get_random_example()
     
     context_desc = example.get('processed_context_desc', '')
@@ -117,10 +111,10 @@ def load_context(set_interrupt=False):
     
     return [
         example,
-        gr.update(value=example['question'], elem_classes="query-text"),  # Regular query styles
+        gr.update(value=example['question'], elem_classes="query-text"),
         gr.update(value=context_desc, visible=bool(context_desc)),
         gr.update(value=context_html),
-        gr.update(value="Show Full Context", elem_classes=["context-toggle-button"], visible=True),  # Ensure toggle is visible
+        gr.update(value="Show Full Context", elem_classes=["context-toggle-button"], visible=True),
         show_full
     ]
 
@@ -149,13 +143,6 @@ def generate_model_summaries(example):
         "completed": False
     }
     
-    # Clear the interrupt flag when NEW inference starts
-    generation_interrupt.clear()
-    print("Interrupt flag cleared for new inference")
-    
-    if generation_interrupt.is_set():
-        return result
-
     try:
         # Get current leaderboard data to determine model usage counts
         leaderboard_data = load_leaderboard_data()
@@ -188,13 +175,10 @@ def generate_model_summaries(example):
         print(f"Starting generation with models: {m_a_name} vs {m_b_name}")
         s_a, s_b = generate_summaries(example, m_a_name, m_b_name)
         
-        if not generation_interrupt.is_set():
-            result["summary_a"] = s_a
-            result["summary_b"] = s_b
-            result["completed"] = bool(s_a and s_b)
-            print("Generation completed successfully")
-        else:
-            print("Generation was interrupted")
+        result["summary_a"] = s_a
+        result["summary_b"] = s_b
+        result["completed"] = bool(s_a and s_b)
+        print("Generation completed successfully")
     except Exception as e:
         print(f"Error in generation: {e}")
         
@@ -208,8 +192,8 @@ def process_generation_result(result):
             result.get("summary_a", ""), 
             result.get("summary_b", ""),
             None, [], False, load_leaderboard_data(),
-            gr.update(value=result.get("summary_a", "Generation was interrupted or failed.")),
-            gr.update(value=result.get("summary_b", "Generation was interrupted or failed.")),
+            gr.update(value=result.get("summary_a", "Generation failed or was interrupted.")),
+            gr.update(value=result.get("summary_b", "Generation failed or was interrupted.")),
             gr.update(interactive=False, elem_classes=["vote-button"]),
             gr.update(interactive=False, elem_classes=["vote-button"]),
             gr.update(interactive=False, elem_classes=["vote-button"]),
@@ -218,7 +202,7 @@ def process_generation_result(result):
             gr.update(visible=False),
             gr.update(interactive=False, visible=True),
             gr.update(visible=False),
-            gr.update(interactive=True),
+            gr.update(interactive=True),  # Re-enable button
             gr.update(elem_classes=[])
         ]
     
@@ -239,7 +223,7 @@ def process_generation_result(result):
         gr.update(visible=False),
         gr.update(interactive=False, visible=True),
         gr.update(visible=False),
-        gr.update(interactive=True),
+        gr.update(interactive=True),  # Re-enable button
         gr.update(elem_classes=[])
     ]
 
@@ -295,12 +279,12 @@ def show_loading_state():
         gr.update(visible=False),      # feedback_section
         gr.update(interactive=False),  # submit_button
         gr.update(visible=False),      # results_reveal_area
-        gr.update(interactive=True),   # random_question_btn - KEEP ACTIVE during inference
+        gr.update(interactive=False),  # DISABLE button during inference
         None  # Reset selected_winner
     ]
 
 def handle_new_example_click():
-    return load_context(set_interrupt=True)[0]
+    return load_context()[0]
 
 def update_ui_for_new_context(example):
     context_desc = example.get('processed_context_desc', '')
@@ -308,10 +292,10 @@ def update_ui_for_new_context(example):
         context_desc = f"<div class='context-topic'><span class='topic-label'>The question and context are about:</span> {context_desc}</div>"
     
     return [
-        gr.update(value=example['question'], elem_classes="query-text"),  # Regular query styles
+        gr.update(value=example['question'], elem_classes="query-text"),
         gr.update(value=context_desc, visible=bool(context_desc)),
         gr.update(value=get_context_html(example, False)),
-        gr.update(value="Show Full Context", elem_classes=["context-toggle-button"], visible=True),  # Ensure toggle is visible
+        gr.update(value="Show Full Context", elem_classes=["context-toggle-button"], visible=True),
         False
     ]
 
@@ -326,7 +310,7 @@ def reset_reference_section():
 
 def cleanup_on_disconnect():
     print(f"Browser disconnected. Cleaning up resources...")
-    generation_interrupt.set()
+    # Remove interrupt logic
 
 # Helper functions for showing/hiding UI elements
 def initialize_empty_app():
@@ -343,7 +327,7 @@ def show_all_after_loading():
         gr.update(visible=True),  # model_section
         gr.update(visible=True),  # voting_section
         gr.update(visible=True),  # submit_button
-        gr.update(value="🔄 Try a New Question", elem_classes=["query-button"], interactive=True)  # update button text and re-enable
+        gr.update(value="🔄 Try a New Question", elem_classes=["query-button"], interactive=True)  # RE-ENABLE button
     ]
 
 with gr.Blocks(theme=gr.themes.Default(
@@ -377,8 +361,8 @@ with gr.Blocks(theme=gr.themes.Default(
     show_results_state = gr.State(False)
     results_agg = gr.State(load_leaderboard_data())
     show_full_context = gr.State(False)
-    show_reference_answer = gr.State(False)  # NEW: State for reference answer toggle
-    faq_expanded = gr.State(False)  # State for FAQ toggle
+    show_reference_answer = gr.State(False)
+    faq_expanded = gr.State(False)
 
     with gr.Tabs() as tabs:
         with gr.TabItem("Arena", id="arena-tab"):
@@ -430,7 +414,6 @@ with gr.Blocks(theme=gr.themes.Default(
                 with gr.Column(visible=False, elem_id="model-section") as model_section:
                     gr.Markdown("---")
                     
-                    # NEW: Model comparison header (simple)
                     gr.Markdown("### 🔍 Compare Models - Are these Grounded, Complete Answers or Correct Rejections?", elem_classes="section-heading")
 
                     with gr.Row(elem_id="summary-containers"):
@@ -455,12 +438,11 @@ with gr.Blocks(theme=gr.themes.Default(
                                     elem_id="summary-b-display"
                                 )
                     
-                    
-                    # NEW: Reference Answer Toggle (exactly like FAQ style)
+                    # Reference Answer Toggle
                     with gr.Row(elem_id="reference-toggle-row"):
                         reference_toggle_btn = gr.Button("▶ Show Reference Answer", elem_classes=["faq-toggle-button"])
                     
-                    # Reference Answer Content - initially hidden (exactly like FAQ)
+                    # Reference Answer Content - initially hidden
                     with gr.Row(visible=False, elem_id="reference-content") as reference_content:
                         reference_answer_display = gr.Markdown("", elem_classes="faq-text")
 
@@ -527,7 +509,7 @@ The Elo rating system provides a more accurate ranking than simple win rates:
         outputs=[show_full_context, context_display, context_toggle_btn]
     )
     
-    # NEW: Reference answer toggle functionality (exactly like FAQ)
+    # Reference answer toggle functionality
     reference_toggle_btn.click(
         fn=toggle_reference_answer,
         inputs=[show_reference_answer, current_example],
@@ -573,12 +555,10 @@ The Elo rating system provides a more accurate ranking than simple win rates:
         outputs=[query_display, context_description, context_display, 
                 context_toggle_btn, show_full_context]
     ).then(
-        # NEW: Reset reference section when loading new question
         fn=reset_reference_section,
         inputs=[],
         outputs=[show_reference_answer, reference_content, reference_toggle_btn, reference_answer_display]
     ).then(
-        # IMPORTANT: Explicitly hide FAQ here
         fn=hide_faq_section,
         inputs=[],
         outputs=[faq_container, faq_content]
@@ -622,12 +602,10 @@ The Elo rating system provides a more accurate ranking than simple win rates:
         outputs=[query_display, context_description, context_display, 
                 context_toggle_btn, show_full_context]
     ).then(
-        # NEW: Reset reference section when trying another question
         fn=reset_reference_section,
         inputs=[],
         outputs=[show_reference_answer, reference_content, reference_toggle_btn, reference_answer_display]
     ).then(
-        # IMPORTANT: Explicitly hide FAQ here too
         fn=hide_faq_section,
         inputs=[],
         outputs=[faq_container, faq_content]