Add gradio logging

app.py

@@ -4,7 +4,7 @@ import json
 import uuid
 import datetime
 import os
-from typing import Dict, List, Any, Tuple
+from typing import Dict, List, Any, Tuple, Optional
 
 # Import our modules
 import data_utils
@@ -32,7 +32,7 @@ def load_example(state):
         current_output = data_utils.all_outputs[state["current_example_index"]]
         
         # Return values in the same order as the outputs parameter in demo.load
-        return current_output["low_inference_notes"], current_output["output"]["summary"]
+        return current_output["component_id"], current_output["low_inference_notes"], current_output["output"]["summary"]
     except (IndexError, KeyError) as e:
         # Handle errors by initializing a new cycle
         data_utils.initialize_cycle()
@@ -40,11 +40,12 @@ def load_example(state):
         current_output = data_utils.all_outputs[0]
         
         # Return values in the same order as the outputs parameter in demo.load
-        return current_output["low_inference_notes"], current_output["output"]["summary"]
+        return current_output["component_id"], current_output["low_inference_notes"], current_output["output"]["summary"]
 
 def select_choice(choice, state):
-    """Just set the choice in the state without submitting"""
+    """Set the choice in the state and provide visual feedback"""
     state["current_choice"] = choice
+    print(f"User selected: {choice}")
     return {
         "good_btn": gr.update(variant="primary" if choice == "good" else "secondary", 
                             elem_classes=["good-button", "selected-button"] if choice == "good" else ["good-button"]),
@@ -52,6 +53,7 @@ def select_choice(choice, state):
                             elem_classes=["fine-button", "selected-button"] if choice == "fine" else ["fine-button"]),
         "bad_btn": gr.update(variant="primary" if choice == "bad" else "secondary",
                             elem_classes=["bad-button", "selected-button"] if choice == "bad" else ["bad-button"]),
+        "log_message": gr.update(value=f"Selected rating: {choice.upper()}", visible=True),
         "state": state
     }
 
@@ -61,26 +63,39 @@ def stage1_submit(explanation, state):
     if not state["current_choice"]:
         return {
             "error_message": gr.update(value="Please select a rating (Good, Fine, or Bad).", visible=True),
+            "log_message": gr.update(value="Submission failed: No rating selected", visible=True),
             "state": state
         }
     
     if not explanation.strip():
         return {
             "error_message": gr.update(value="Please provide an explanation for your choice.", visible=True),
+            "log_message": gr.update(value="Submission failed: No explanation provided", visible=True),
             "state": state
         }
     
     # Get current output
     current_output = data_utils.all_outputs[state["current_example_index"]]
     
+    # Log the action
+    print(f"Submitting feedback for output ID: {current_output['id']}")
+    print(f"Choice: {state['current_choice']}, Explanation: {explanation}")
+    
     # Store feedback in Argilla
-    data_utils.store_feedback(
+    success = data_utils.store_feedback(
         output_id=current_output["id"],
         choice=state["current_choice"],
         explanation=explanation,
         stage="individual"
     )
     
+    if not success:
+        return {
+            "error_message": gr.update(value="Failed to store feedback in Argilla.", visible=True),
+            "log_message": gr.update(value="Submission failed: Error storing in Argilla", visible=True),
+            "state": state
+        }
+    
     # Increment counter
     state["examples_completed"] += 1
     
@@ -106,10 +121,12 @@ def stage1_submit(explanation, state):
             return {
                 "stage1_group": gr.update(visible=True),
                 "stage2_group": gr.update(visible=False),
+                "component_id_display": current_output["component_id"],
                 "notes_display": current_output["low_inference_notes"],
                 "output_display": current_output["output"]["summary"],
                 "explanation_input": gr.update(value=""),
                 "error_message": gr.update(value="", visible=False),
+                "log_message": gr.update(value="Started new cycle - not enough rated outputs for comparison", visible=True),
                 "good_btn": gr.update(variant="secondary"),
                 "fine_btn": gr.update(variant="secondary"),
                 "bad_btn": gr.update(variant="secondary"),
@@ -127,6 +144,7 @@ def stage1_submit(explanation, state):
             "explanation_input": gr.update(value=""),
             "comparison_explanation": gr.update(value=""),
             "error_message": gr.update(value="", visible=False),
+            "log_message": gr.update(value="Moving to comparison stage", visible=True),
             "good_btn": gr.update(variant="secondary"),
             "fine_btn": gr.update(variant="secondary"),
             "bad_btn": gr.update(variant="secondary"),
@@ -140,10 +158,12 @@ def stage1_submit(explanation, state):
         return {
             "stage1_group": gr.update(visible=True),
             "stage2_group": gr.update(visible=False),
+            "component_id_display": current_output["component_id"],
             "notes_display": current_output["low_inference_notes"],
             "output_display": current_output["output"]["summary"],
             "explanation_input": gr.update(value=""),
             "error_message": gr.update(value="", visible=False),
+            "log_message": gr.update(value=f"Submission successful! Moving to next example ({state['examples_completed']+1}/{len(data_utils.all_outputs)})", visible=True),
             "good_btn": gr.update(variant="secondary"),
             "fine_btn": gr.update(variant="secondary"),
             "bad_btn": gr.update(variant="secondary"),
@@ -151,13 +171,15 @@ def stage1_submit(explanation, state):
         }
 
 def select_comparison_choice(choice, state):
-    """Just set the comparison choice in the state without submitting"""
+    """Set the comparison choice in the state and provide visual feedback"""
     state["current_choice"] = choice
+    print(f"User selected output: {choice}")
     return {
         "choose_a_btn": gr.update(variant="primary" if choice == "a" else "secondary",
                                 elem_classes=["selected-button"] if choice == "a" else []),
         "choose_b_btn": gr.update(variant="primary" if choice == "b" else "secondary",
                                 elem_classes=["selected-button"] if choice == "b" else []),
+        "log_message": gr.update(value=f"Selected output: {choice.upper()}", visible=True),
         "state": state
     }
 
@@ -167,12 +189,14 @@ def stage2_submit(explanation, state):
     if not state["current_choice"]:
         return {
             "error_message": gr.update(value="Please select which output is better (A or B).", visible=True),
+            "log_message": gr.update(value="Submission failed: No output selected", visible=True),
             "state": state
         }
     
     if not explanation.strip():
         return {
             "error_message": gr.update(value="Please provide an explanation for your choice.", visible=True),
+            "log_message": gr.update(value="Submission failed: No explanation provided", visible=True),
             "state": state
         }
     
@@ -183,17 +207,35 @@ def stage2_submit(explanation, state):
     # Determine which output was chosen
     a_wins = (state["current_choice"] == "a")
     
+    # Log the action
+    print(f"Comparing outputs: {output_a['id']} vs {output_b['id']}")
+    print(f"User chose: {'A' if a_wins else 'B'}, Explanation: {explanation}")
+    
     # Update ELO ratings
-    elo.update_ratings(output_a["id"], output_b["id"], a_wins)
+    new_rating_a, new_rating_b = elo.update_ratings(output_a["id"], output_b["id"], a_wins)
+    
+    if new_rating_a is None or new_rating_b is None:
+        return {
+            "error_message": gr.update(value="Failed to update ELO ratings.", visible=True),
+            "log_message": gr.update(value="Submission failed: Error updating ELO ratings", visible=True),
+            "state": state
+        }
     
     # Store feedback
-    data_utils.store_feedback(
+    success = data_utils.store_feedback(
         output_id=output_a["id"] if a_wins else output_b["id"],
         choice=f"preferred_over_{output_b['id'] if a_wins else output_a['id']}",
         explanation=explanation,
         stage="comparison"
     )
     
+    if not success:
+        return {
+            "error_message": gr.update(value="Failed to store comparison feedback in Argilla.", visible=True),
+            "log_message": gr.update(value="Submission failed: Error storing comparison in Argilla", visible=True),
+            "state": state
+        }
+    
     # Reset current choice
     state["current_choice"] = None
     
@@ -210,11 +252,13 @@ def stage2_submit(explanation, state):
         return {
             "stage1_group": gr.update(visible=True),
             "stage2_group": gr.update(visible=False),
+            "component_id_display": current_output["component_id"],
             "notes_display": current_output["low_inference_notes"],
             "output_display": current_output["output"]["summary"],
             "explanation_input": gr.update(value=""),
             "comparison_explanation": gr.update(value=""),
             "error_message": gr.update(value="", visible=False),
+            "log_message": gr.update(value="Completed comparison cycle. Starting new evaluation cycle.", visible=True),
             "choose_a_btn": gr.update(variant="secondary"),
             "choose_b_btn": gr.update(variant="secondary"),
             "state": state
@@ -232,11 +276,13 @@ def stage2_submit(explanation, state):
             return {
                 "stage1_group": gr.update(visible=True),
                 "stage2_group": gr.update(visible=False),
+                "component_id_display": current_output["component_id"],
                 "notes_display": current_output["low_inference_notes"],
                 "output_display": current_output["output"]["summary"],
                 "explanation_input": gr.update(value=""),
                 "comparison_explanation": gr.update(value=""),
                 "error_message": gr.update(value="", visible=False),
+                "log_message": gr.update(value="Not enough rated outputs for comparison. Starting new cycle.", visible=True),
                 "choose_a_btn": gr.update(variant="secondary"),
                 "choose_b_btn": gr.update(variant="secondary"),
                 "state": state
@@ -252,6 +298,7 @@ def stage2_submit(explanation, state):
             "output_b": output_b["output"]["summary"],
             "comparison_explanation": gr.update(value=""),
             "error_message": gr.update(value="", visible=False),
+            "log_message": gr.update(value=f"Comparison submitted successfully! Moving to next comparison ({state['elo_comparisons_done']+1}/5)", visible=True),
             "choose_a_btn": gr.update(variant="secondary"),
             "choose_b_btn": gr.update(variant="secondary"),
             "state": state
@@ -302,12 +349,20 @@ with gr.Blocks(theme=gr.themes.Soft(
                    opacity: 0.8;
                    box-shadow: 0 0 5px rgba(0,0,0,0.5);
                }
+               .log-message {
+                   padding: 8px;
+                   margin-top: 10px;
+                   border-radius: 4px;
+                   background-color: #f0f0f0;
+                   font-size: 14px;
+               }
                """) as demo:
     # Initialize state
     state = gr.State(initialize_state())
     
     # Error message area
     error_message = gr.Markdown(visible=False)
+    log_message = gr.Markdown(visible=False, elem_classes=["log-message"])
     
     with gr.Tabs():
         with gr.TabItem("Data Labeling"):
@@ -316,6 +371,9 @@ with gr.Blocks(theme=gr.themes.Soft(
                 gr.Markdown("# Output Evaluation")
                 gr.Markdown("### Please rate the quality of this output")
                 
+                # Component ID display
+                component_id_display = gr.Markdown("### Component ID: Loading...")
+                
                 with gr.Row():
                     # Left column for notes - make it scrollable and smaller
                     with gr.Column(scale=1):
@@ -374,7 +432,7 @@ with gr.Blocks(theme=gr.themes.Soft(
                             label="Output A",
                             lines=8,
                             interactive=False
-                        )
+                                                    )
                         choose_a_btn = gr.Button("Output A is better", variant="secondary")
                     
                     with gr.Column():
@@ -402,7 +460,7 @@ with gr.Blocks(theme=gr.themes.Soft(
     demo.load(
         fn=load_example,
         inputs=[state],
-        outputs=[notes_display, output_display]
+        outputs=[component_id_display, notes_display, output_display]
     )
     
     # Also load the leaderboard on startup
@@ -411,6 +469,62 @@ with gr.Blocks(theme=gr.themes.Soft(
         inputs=[],
         outputs=[leaderboard_display]
     )
+    
+    # Connect the buttons to their functions
+    good_btn.click(
+        fn=select_choice,
+        inputs=["good", state],
+        outputs=[good_btn, fine_btn, bad_btn, log_message, state]
+    )
+    
+    fine_btn.click(
+        fn=select_choice,
+        inputs=["fine", state],
+        outputs=[good_btn, fine_btn, bad_btn, log_message, state]
+    )
+    
+    bad_btn.click(
+        fn=select_choice,
+        inputs=["bad", state],
+        outputs=[good_btn, fine_btn, bad_btn, log_message, state]
+    )
+    
+    submit_btn.click(
+        fn=stage1_submit,
+        inputs=[explanation_input, state],
+        outputs=[
+            stage1_group, stage2_group, component_id_display, notes_display, output_display,
+            explanation_input, error_message, log_message, good_btn, fine_btn, bad_btn, state
+        ]
+    )
+    
+    choose_a_btn.click(
+        fn=select_comparison_choice,
+        inputs=["a", state],
+        outputs=[choose_a_btn, choose_b_btn, log_message, state]
+    )
+    
+    choose_b_btn.click(
+        fn=select_comparison_choice,
+        inputs=["b", state],
+        outputs=[choose_a_btn, choose_b_btn, log_message, state]
+    )
+    
+    compare_submit_btn.click(
+        fn=stage2_submit,
+        inputs=[comparison_explanation, state],
+        outputs=[
+            stage1_group, stage2_group, component_id_display, notes_display, output_display,
+            notes_display_elo, output_a, output_b, comparison_explanation,
+            error_message, log_message, choose_a_btn, choose_b_btn, state
+        ]
+    )
+    
+    refresh_btn.click(
+        fn=get_leaderboard_display,
+        inputs=[],
+        outputs=[leaderboard_display]
+    )
 
 # For Hugging Face Spaces deployment
 if __name__ == "__main__":
@@ -418,4 +532,5 @@ if __name__ == "__main__":
     data_utils.init_argilla()
     
     # Launch the app
-    demo.launch()
+    demo.launch()
+