Update app.py

app.py

@@ -30,7 +30,7 @@ logging.basicConfig(filename="chat_log.txt", level=logging.DEBUG, format='%(asct
 chat_queue = Queue()  # Define chat_queue globally
 
 # --- Chat Function with Monitoring ---
-def chat_function(message, ground_truth):
+def chat_function(message, user_ner_tags, ground_truth):
     logging.debug("Starting chat_function")
     with REQUEST_LATENCY.time():
         REQUEST_COUNT.inc()
@@ -42,7 +42,8 @@ def chat_function(message, ground_truth):
             logging.debug(f"NER results: {ner_results}")
 
             detailed_response = []
-            predicted_labels = []
+            model_predicted_labels = []
+            user_predicted_labels = []
             for result in ner_results:
                 token = result['word']
                 score = result['score']
@@ -50,7 +51,7 @@ def chat_function(message, ground_truth):
                 start = result['start']
                 end = result['end']
                 label_id = int(entity.split('_')[-1])  # Extract numeric label from entity
-                predicted_labels.append(label_id)
+                model_predicted_labels.append(label_id)
                 detailed_response.append(f"Token: {token}, Entity: {entity}, Score: {score:.4f}, Start: {start}, End: {end}")
 
             response = "\n".join(detailed_response)
@@ -61,15 +62,21 @@ def chat_function(message, ground_truth):
 
             time.sleep(random.uniform(0.5, 2.5))  # Simulate processing time
 
-            # Compute metrics
+            # Compare user's input tags to the model's output
             try:
-                ground_truth_labels = json.loads(ground_truth)  # Assuming ground_truth is input as a JSON string
+                user_ner_results = json.loads(user_ner_tags)
+                if not isinstance(user_ner_results, list):
+                    raise ValueError("Invalid format for user NER tags. Please provide a JSON list of dictionaries.")
+                for result in user_ner_results:
+                    entity = result['entity']
+                    label_id = int(entity.split('_')[-1])  # Extract numeric label from entity
+                    user_predicted_labels.append(label_id)
             except json.JSONDecodeError:
-                return "Invalid JSON format for ground truth labels. Please provide a valid JSON array."
+                return "Invalid JSON format for user NER tags. Please provide a valid JSON array."
 
-            precision = precision_score(ground_truth_labels, predicted_labels, average='weighted', zero_division=0)
-            recall = recall_score(ground_truth_labels, predicted_labels, average='weighted', zero_division=0)
-            f1 = f1_score(ground_truth_labels, predicted_labels, average='weighted', zero_division=0)
+            precision = precision_score(user_predicted_labels, model_predicted_labels, average='weighted', zero_division=0)
+            recall = recall_score(user_predicted_labels, model_predicted_labels, average='weighted', zero_division=0)
+            f1 = f1_score(user_predicted_labels, model_predicted_labels, average='weighted', zero_division=0)
 
             metrics_response = (f"Precision: {precision:.4f}\n"
                                 f"Recall: {recall:.4f}\n"
@@ -115,9 +122,9 @@ body {
     with gr.Tab("Chat"):
         gr.Markdown("## Chat with the Bot")
         message_input = gr.Textbox(label="Enter your sentence:", lines=2)
-        ground_truth_input = gr.Textbox(label="Enter ground truth labels (JSON format):", lines=2)
+        user_ner_tags_input = gr.Textbox(label="Enter your NER tags (JSON format):", lines=5)
         output = gr.Textbox(label="Response", lines=10)
-        chat_interface = gr.Interface(fn=chat_function, inputs=[message_input, ground_truth_input], outputs=output)
+        chat_interface = gr.Interface(fn=chat_function, inputs=[message_input, user_ner_tags_input, gr.Textbox(lines=5)], outputs=output)
         chat_interface.render()
 
     with gr.Tab("Model Parameters"):
@@ -199,4 +206,4 @@ body {
     threading.Thread(target=update_queue_length, daemon=True).start()
 
 # Launch the app
-demo.launch()
+demo.launch(share=True)